预算辩论 · 2026-03-02 · 第 15 届国会
2026数字发展与信息部供给委员会辩论：AI作为战略优势
MDDI Committee of Supply 2026 — AI as Strategic Advantage
这是2026年预算案中AI议题最集中的辩论。MDDI GPC围绕六大主题进行协调审议：AI价值主张、数字能力建设、伦理治理、包容性增长、基础设施与网络安全、高信任数字社会。部长尤芳达(Mrs Josephine Teo)宣布：(1)支持10万名工人成为"AI双语人才"，从会计和法律行业开始，通过TeSA项目扩展；(2)发布全球首个《自主智能体AI治理框架》(Model Governance Framework for Agentic AI)；(3)新加坡将主办第二届国际AI安全科学交流大会，更新"新加坡共识"；(4)着力解决中小企业AI落差问题，避免大企业拉远距离。议员关注焦点包括：深伪技术监管(Christopher de Souza)、AI媒体素养(Fadli Fawzi)、数据中心投资竞争、AI对PME的冲击、网络安全应对AI威胁等。
关键要点
- • 10万工人"AI双语"计划
- • 全球首个自主智能体AI治理框架
- • 第二届国际AI安全科学交流大会
- • 中小企业AI落差是核心关注
- • 深伪监管与AI媒体素养
- • 网络安全应对AI威胁
积极推动AI民主化，强调中小企业普惠与风险治理并行
He Ting Ru关注AI对弱势群体的影响和公共信息架构的韧性
AI治理从原则走向具体行动，自主智能体框架开全球先河
"Not all of us can be AI engineers. But we can be "bilingual" in AI in our own areas of expertise."
参与人员（11）
完整译文（中文）
Hansard 英文原文译文 · 翻译日期：2026-05-02
主席：数字发展与信息部（MDDI）Q组负责人。沙拉尔·塔哈先生。
下午3时17分
将人工智能作为战略优势
沙拉尔·塔哈先生（巴西立-樟宜选区）：谢谢主席先生。主席先生，我提议，“将预算中Q组的总拨款减少100元”。
主席先生，人工智能（AI）在本预算中占据显著位置，本议院早已认识到其重要性。从2019年的国家人工智能战略到2023年的国家人工智能战略2.0（NAIS 2.0），我们已从试验阶段迈向规模化。MDDI政府议会委员会（GPC）一直在六个关键主题上积极推动。我们今天的削减反映了协调一致的审查。
首先，加强新加坡的人工智能价值主张。我的削减旨在澄清我们的全球竞争优势以及国家人工智能理事会的角色，而朱佩玲博士和陈洁仪女士将推动实现可衡量的成果，而不仅仅是技术活动。
第二，建设深厚且广泛的数字能力。郭振辉先生、朱佩玲博士、李嘉欣女士和我将就如何在劳动力中推广人工智能技能，以及如何帮助跨国企业（MNEs）和中小企业（SMEs）有意义地采用人工智能提出削减建议。
第三，确保伦理数字治理。苏世民先生、陈洁仪女士、田佩玲女士和我将就监管保障和问责框架提出削减建议，以确保伦理和负责任的增长，尤其是在我们迈向更自主、具代理性的人工智能系统时。
第四，通过技术实现包容性增长并提升弱势群体。李嘉欣女士和我将就为应届毕业生、青年、长者和低收入群体创造机会提出削减建议，以便技术能够扩大机会。
第五，投资基础设施和网络安全。陈洁仪女士和我将审视我们如何在日益复杂的人工智能驱动威胁中加强网络安全态势。
第六，建设高信任的数字社会。田佩玲女士和陈洁仪女士将就信任、安全以及防范诈骗、深度伪造和网络危害提出削减建议。
这些削减反映了我们GPC的审慎、一致和坚持，与部委共同推动工作，保障新加坡的数字未来。
主席先生，请允许我从第一个削减开始，关于加强新加坡在全球人工智能领域的价值主张及保持全球市场相关性。
主席先生，全球人工智能竞赛急剧加速。在最近的全球人工智能峰会上，显然我们已进入由大规模计算和多模态能力驱动的前沿和基础模型时代。
生成式人工智能（GenAI）不再是实验性质。它已嵌入企业、公共服务和国家系统。向具代理性的人工智能和人工智能原生企业的转变标志着结构性变革。
这种变革在三个层面展开。首先，人口规模。在中国，人工智能整合于服务数亿人的平台。在美国，人工智能助手嵌入全球使用的生产力工具中。在印度，人工智能融入国家规模的电信和数字服务。人工智能已成为日常工作流程的一部分。
第二，计算规模。竞赛不仅仅是计算主导权。下一代芯片正以前所未有的数量被订购。超级规模云服务商在数据中心投资数十亿美元。印度宣布吸引高达2000亿美元的人工智能和数据中心投资的雄心，中东和欧盟则在确保主权计算能力。芯片、数据中心和能源现已成为战略基础设施。
第三，产业规模。人工智能嵌入制造、物流、国防和能源系统。这关乎产业竞争力和国家能力。随着能力加速，责任必须同步。安全对齐评估和红队测试至关重要。信任将决定谁能实现规模化。
新加坡无法在人口规模或计算规模上竞争，但我们可以在精准度、信任、监管信誉和深度行业整合上竞争。这不仅关乎技术，更关乎就业和国家竞争力。
新加坡在全球人工智能格局中的独特价值主张是什么？我们如何与美国、中国和印度等人工智能巨头的人口规模、计算规模和产业规模竞争？我们如何利用我们的高信任治理、监管信誉和深度行业集中度？
因此，我欢迎国家人工智能理事会的成立。
为了成功，明确的授权和执行能力至关重要。它的具体角色是什么？它是否拥有执行权，监督跨部委实施，还是仅作为咨询机构？
在快速发展的人工智能竞赛中，结构必须引导果断行动。理事会如何整合贸易与工业部（MTI）下的经济战略与MDDI下的数字治理，以确保协调交付？
国家人工智能战略2.0（NAIS 2.0）于2023年发布。取得了哪些进展？理事会会在此基础上加强，而非重复或削弱现有努力吗？当优先事项发生冲突时，如何解决跨机构摩擦？
最后，理事会如何保持与产业现实的联系？产业领袖会支持并成为理事会成员吗？中小企业会有实质性发言权吗？
因此，我支持集中资源聚焦国家人工智能使命，并有人工智能的推动者加速部署。但我希望向部委寻求澄清。
政府如何确定国家人工智能使命的四个关键行业？使用了哪些标准？这些使命的成功如何定义？必须超越试点，达到可衡量的真实经济和社会成果。随着步伐加快，我们何时考虑扩展更多使命？我希望是在几个月内。
最后，我们如何定义和衡量人工智能推动者在能力、采用、全球竞争力以及对就业和生产力的实际影响方面的成功？
最后一点。为了在全球人工智能供应链中保持相关性，我们必须为计算基础设施做好准备。
我们发展更多计算能力的计划是什么？鉴于地缘政治紧张局势，我们不能仅依赖海外支持的计算需求。我们拥有多少主权人工智能计算能力？如何减少对他人计算能力的依赖？此外，我们支持人工智能规模化数据中心的清洁或可再生能源战略是什么？
（程序文本）提案问题。（程序文本）
主席：沙拉尔·塔哈先生。您可以将三个削减一并提出。
将人工智能作为战略优势
沙拉尔·塔哈先生：谢谢主席先生。如果新加坡要将人工智能作为战略优势，必须同步推进三大推动力——深厚且广泛的劳动力能力、广泛的企业采用（包括中小企业）以及强有力的伦理治理。政府已推出多项资助计划，包括首席技术官（CTO）即服务。我们在加速人工智能转型方面取得了哪些可衡量的进展？存在哪些差距？
（副议长谢耀权先生主持）
释放人工智能价值不仅仅需要工具，还需要重新设计业务流程和重新定义运营模式。我们如何帮助企业将人工智能整合到核心工作流程，打造人工智能准备团队，并实现切实的生产力提升？
对于中小企业，存在三大制约因素：人才成本、缺乏验证的用例和集成复杂性。如果人工智能采用仅集中于大型企业，中小企业的生产力和工资差距将扩大。我们必须从咨询支持转向共享能力基础设施。
因此，我提出三项改进建议。
第一，将我们的CTO即服务发展为人工智能能力即服务，派遣集中调配的人工智能工程师团队，跨中小企业集群部署解决方案，进行实操实施。第二，开发共享行业平台，共同创建即插即用的人工智能模型，用于质量控制和物流优化等通用功能，帮助降低实验成本，特别是对中小企业。第三，引入与成果挂钩的共同资助，关联可衡量的生产力、出口增长或能源效率成果。人工智能基础设施应被视为共享的物理资源，使我们的中小企业能够通过敏捷性而非规模竞争。
除了企业之外，劳动力发展也至关重要。通过诸如TeSA等项目，我们在培养深度人工智能专业知识方面取得了哪些进展？在提升更广泛劳动力的人工智能素养方面又有哪些成效？人工智能如何为老年人、应届毕业生以及重返职场的女性拓宽机会？我们能否为老年人创造结构化路径，使他们能够有意义地使用人工智能工具？随着入门级岗位被颠覆，我们如何重新设计工作和学徒计划，使毕业生既获得技术技能，也具备商业背景知识？
最后，随着具备自主性的智能代理系统日益增多，我们如何确保其伦理和安全部署？
人工智能在伦理上明显有益的领域有很多——例如协助医生诊断、检测金融欺诈、优化能源消耗或支持老年人日常生活。但也有不可逾越的界限，比如国防领域的自主致命决策、操控性行为定向、不透明的信用评分加剧偏见，或人工智能代理在无责任追究的情况下做出就业决策。
当智能代理系统以有限透明度运行，目标不断演变，甚至出现开发者难以完全预测的新兴行为时，挑战更加复杂。发生损害时，最终责任由谁承担？是开发者、部署者还是操作员？将强制实施何种治理框架、审计要求、红队测试标准和可解释性门槛，以确保信任与能力同步提升？
数字公益
主席先生，随着人工智能和数字化转型的推进，我们如何确保技术真正成为一股向善的力量？
首先，政府如何加强关键服务的数字化交付，特别是针对视觉障碍等特殊需求人士？我们的系统是否从设计上包容多样性？其次，随着社会老龄化，数字和人工智能解决方案如何更好地支持老年人，同时减轻“夹心层”家庭的负担？第三，如何确保低收入家庭的儿童不仅能获得设备，还能掌握人工智能技能并获得相关机会？最后，如何加强数字空间的信任，更好地保护新加坡人免受诈骗和网络伤害？
技术不能加剧分化，必须提升、保护并赋能每一位新加坡人。
加强我们的网络安全态势
网络威胁不再是孤立事件。它们持续存在、不断适应，且越来越多地借助人工智能。随着人工智能系统变得更具自主性，我们必须面对一类新风险——能够独立规划、侦察和行动的人工智能代理。
MDDI如何应对恶意行为者利用人工智能自动化侦察、策划复杂钓鱼攻击或大规模利用漏洞的威胁？
下午3点30分
新加坡曾遭遇过安全漏洞。2018年SingHealth网络攻击事件泄露了150万患者数据。最近有报道称，与新加坡关键基础设施相关的255家公司遭到针对性攻击，勒索软件事件扰乱了医疗集群和第三方供应商。去年，一起影响关键基础信息基础设施（CII）运营商的网络事件提醒我们，从能源到交通，我们的关键基础设施仍是主要攻击目标。
高级持续威胁（APT）耐心且资源充足，可能与国家有关联。他们不仅寻求破坏，更谋求战略杠杆。那么，我们如何加强国家网络安全态势以防御APT？我们是否在威胁情报融合、实时监控和跨部门事件响应方面投入充足？
关于关键基础设施，法规必须跟上威胁演变。政府将如何加强对关键基础设施运营商的网络安全要求？除了合规之外，将提供哪些工具、共享平台和人工智能驱动的检测能力，帮助运营商防御复杂攻击？
同样重要的是，鉴于供应链脆弱性，我们如何加强不仅关键基础设施所有者，还包括其供应商和网络安全服务提供商的安全态势？
最后，网络安全归根结底是关于人的。我们如何扩大和深化网络安全人才队伍，以应对当今威胁？我们是否加快了专业培训、中年职业转换和高级人工智能安全集成技能的培养？随着人工智能既成为威胁因素又是防御工具，我们如何支持组织负责任且有效地采用人工智能驱动的网络安全解决方案？
在数字冲突日益升级的世界中，韧性不是可选项。数字经济的信任依赖于我们的防御能力。我们必须确保新加坡大规模数字化的同时，网络安全态势同步增强。
网络防御
严彦松议员（阿裕尼） ：主席先生，2026年2月9日，政府披露新加坡主要电信运营商去年遭到UNC3886组织发起的复杂网络间谍活动攻击。这类入侵清楚地提醒我们，数字战场已扩展为战略破坏的战区，APT行为者预置恶意代码，潜伏多年，设计在危机时刻激活，触发停电或扰乱交通和支付系统。
对新加坡而言，这直接威胁国家生存，因为对民用电信、支付系统和交通网络的协调破坏将直接削弱新加坡武装部队快速动员和部署部队的能力。虽然遏制UNC3886展示了我们的技术能力，但我们必须利用这一能力发出明确后果信号。政府应与国际伙伴合作，传达战略红线，明确表示在关键基础设施中预置恶意代码是不可接受的挑衅。我们必须利用归因能力直接点名此类行为者，同时谨慎权衡点名国家关联组织的外交敏感性。我们应朝着通过精准信号和校准反制威胁实现主动威慑的态势迈进。如此一来，在遵守国际法的前提下，我们可以避免意外升级。最终，我们必须有效改变潜在侵略者的成本收益计算。
量子安全密码学解决方案——为何背离全球共识
张文健议员（阿裕尼） ：主席先生，新加坡的立场与全球共识背道而驰。2024年1月，法国、德国、荷兰和瑞典的网络安全机构联合评估量子密钥分发（QKD），结论是：QKD尚未成熟，仅适用于小众用例。他们表示：“迁移到后量子密码学（PQC）优先于使用QKD。”
PQC优先不仅是美国立场。PQC算法由欧洲研究人员构建，备选算法完全由法国开发。来自25个国家的82个候选算法经过8年公开密码分析。德国于2020年发布PQC迁移指南，早于标准最终确定4年。澳大利亚计划于2030年停止使用经典公钥密码学；日本为2035年；18个欧盟国家去年11月签署了PQC承诺，未提及QKD。PQC是软件，可部署于现有基础设施。苹果通过iOS更新向13亿设备推送PQC，谷歌为34亿Chrome用户启用，Cloudflare自2023年底起保护全球20%网络流量。无需新光纤，无需专用硬件，仅软件更新。
没有PQC，敌手今天收集数据，明天解密。
新加坡立场相反。我们正在扩展国家量子安全网络，配备专用QKD光纤，却没有设定PQC迁移截止日期。只有新加坡和中国将QKD作为国家基础设施扩展，而非视为小众研究试点。我们的旗舰量子衍生企业将QKD技术卖回政府，政府为其提供资金。
我想问，为什么QKD与PQC的平衡与其他可比国家相反？许多新加坡量子研究人员对此持怀疑态度，他们应得到问责。部长是否会披露量子安全预算在QKD和PQC之间的分配情况，以及新加坡何时会设定PQC迁移截止日期？
主席 ：何亭如女士，您可以将三次发言合并进行。
不雅人工智能内容
何庭如女士（盛港选区）：主席先生，即将颁布的《网络安全（救济与问责）法案》加强了对不雅网络内容受害者的支持。信息通信媒体发展局（IMDA）一直在与X平台就Grok生成未经同意的亲密图像并在X平台大量传播一事进行接触。IMDA表示，X已采取措施解决该问题，包括阻止Grok制作此类内容。即使我们确保科技平台的运营环境不过于限制，政府能否进一步说明与X接触的结果？对此事是否采取了任何惩罚性措施？自从推出“辣味模式”功能后，Grok今年一月跃升至新加坡苹果应用商店免费应用前25名。
其次，我在去年的供应委员会（COS）发言中引用过，有报道称学生生成了同学的深度伪造新闻，并在WhatsApp群组中分享。因此，我们必须解决真正的问题。即现有且日益增长的对性化图像的需求，而这种需求因可访问性而加剧。
鉴于大多数受害者是女性和儿童，增加的可访问性对这些群体施加了更大压力。我们必须做更多工作来教育青少年如何使用人工智能，尤其是在小学四年级就开始接触人工智能。鉴于对儿童如何处理人工智能的担忧，教育部（MOE）的性教育方法和人工智能框架如何明确涵盖这一问题？教育部如何调节学生与人工智能聊天机器人的情感互动？
这些图像与新加坡年轻人发展的关系尤为相关，因为平台正努力变得更具吸引力。除了现有立法外，可能还会出现更多关注点，例如不涉及特定受害者但具有社会关注的内容，如人工智能生成的儿童色情内容。
社交媒体与儿童
主席先生，孩子在睡觉时间还在无休止地刷屏，并非出于选择，而是在回应一个设计得几乎无法停止的系统。现行的年龄认证评估、《网络安全（救济与问责）法案》和网络安全行为准则代表了保护儿童网络安全的共同努力。今天，我想问的是，这些措施是否解决了一个尚未解决的区别——内容伤害与设计伤害的区别。
新加坡对此已有认识。圣淘沙和滨海湾金沙赌场的监管通过入场费、排除令和访问限制等措施建立了刻意的阻力。这认识到需要行为设计的干预，而不仅仅是提供更好的风险信息。在社交媒体平台上，无限滚动、自动播放视频和算法推荐是吸引注意力的“黑暗模式”，旨在通过利用寻求奖励的心理并削弱大脑尚在发育的儿童的自我调节能力来最大化参与度。
上个月，欧盟委员会初步认定TikTok的成瘾性设计本身构成法律违规。TikTok被纳入我们自己的行为准则，委员会发现其屏幕时间工具和家长控制未能有效应对这些风险。新加坡对此保持沉默，增加了声誉风险。上周《自然健康》杂志发表文章指出，我们必须让平台对其成瘾性设计负责。这些平台利用儿童的大脑，削弱儿童的自我调节能力。因此，问题是我们是否应允许平台对儿童部署吸引注意力的“黑暗模式”而不承担法律后果。
部长能否澄清三点？一、行为准则是否要求指定服务提交设计风险评估，涵盖推荐系统、自动播放和滚动架构？IMDA是否有权独立于内容分类对这些评估采取行动？如果有，我们是否会承诺一个时间表？
二、鉴于对TikTok成瘾性设计的发现，IMDA是否已据此审查TikTok的合规报告？
三、教育部是否会考虑我上周提出的建议，成立特别委员会更好地审视全球保护儿童免受社交媒体伤害的努力，尤其是在全面禁止儿童使用社交媒体的呼声日益高涨的背景下？儿童及其家长应获得一个框架，使平台不仅对其展示的内容负责，也对其构建方式负责。数字环境不会自行形成，它们是被设计的，而设计若不受监管，便成为默认政策。
信天翁档案的教训
主席先生，历史不是单一维度的。它不断等待通过获取更多信息而得到补充。对过去的了解对塑造我们当前的思考和行动方式具有重要意义。解密对这一过程至关重要。当错误信息、误导信息、混淆和不确定性出现时，更可信且可独立验证的信息尤为关键。透明度不仅仅是为了透明而透明。
最近获得的信天翁档案强调了与马来西亚分离是双方协议的事实。这使我们有可能超越围绕被逐出新加坡的创伤历史叙述，推动与我们最亲近邻国的关系发展。
在其他地方，公开爱泼斯坦档案使世界上一些最富有和最有权势的人被追究不当行为责任，并导致安德鲁·蒙巴顿·温莎和彼得·曼德尔森等人被逮捕。无论问题多么严重或平凡，访问和问责在相关人员仍在世时尤为重要。
张玉娟部长表示，在决定是否开放公共档案时，国家机构会考虑，引用其话：“支持对我们集体过去的研究，同时保护敏感信息并遵守相关保密及其他义务。”
我们应增加时间表，要求国家机构和政治权力接受公众问责，避免混淆和误传。我们承诺追求民主。在民主制度中，国家行为需要对其服务的公众及其资金来源负责。公开无法辩护的立场和行为不应被采取。因此，这些决定必须随时准备接受公众审查。知晓这一可能性会促使更大的谨慎和责任感。
主席：法兹利·法乌兹先生，请将您的两段发言合并。
人工智能与媒体素养
法兹利·法乌兹先生（亚历山大选区）：主席先生，随着2026年预算推动新加坡的人工智能雄心，我们必须面对越来越多的新加坡人以前所未有的规模和速度接触到人工智能生成的虚假信息和人工智能驱动的诈骗。2月5日《联合早报》报道了一波声称总理黄循财被迫下台及激烈内部权力斗争的视频激增。
这些视频完全由人工智能在几分钟内生成，据称每个20分钟视频成本仅为1至2美元。MDDI承认已观察到多个网络账户发布关于新加坡国内政治的此类捏造言论。MDDI发言人向《早报》表示，已推出公众教育措施和资源，敦促公众依赖官方渠道，避免分享未经核实的内容。
我欢迎这一回应，但我想知道这些措施是否足够。鉴于人工智能生成虚假信息的规模和复杂性，为什么不对这些视频背后的人使用《防止网络虚假信息和操纵法》（POFMA）？仅靠POFMA等执法工具也无法使社会免疫于虚假信息。我们需要一个能够质疑、核实和批判性评估网络信息的民众。
教育部将开发哪些结构化的长期项目来加强媒体素养和批判性思维，尤其是在老年人等弱势群体中？我们是否会扩大社区工作坊、学校课程和公共宣传，教授公民实用的核实步骤，如检查原始视频、审查来源和咨询权威渠道？我们能否利用人工智能本身来帮助大规模筛选和标记可疑内容？
如果人工智能将欺骗成本降至每个视频1美元，那么不作为的代价可能更高。我们的国家人工智能战略将如何确保新加坡人有能力在日益污染的信息生态中辨别事实与虚构？
解密与国家历史
先生，最近解密的《信天翁档案》改变了我们对分离的理解。几十年来，围绕新加坡独立的官方叙述是，我们被联邦政府突然且单方面地驱逐出马来西亚。这个故事塑造了几代新加坡人对我们国家建国的理解。
然而，《信天翁档案》与流行的叙述有所不同。文件显示，在1964年7月种族骚乱之后，人民行动党与马来西亚联盟党之间已经开始了关于马来西亚内部可能的宪法调整的机密谈判。这些讨论最终导致了分离。
下午3点45分
这一发现并不削弱我们的历史，它表明历史往往比我们想象的更复杂，并丰富了我们对历史的理解。但为什么这一具有历史意义的资料迟迟未被公开？还有多少重要记录仍然无法获取？
最近解密的《信天翁档案》说明了为何需要《信息自由法》和自动解密制度。工人党长期呼吁制定《信息自由法》，最近一次是在我们2025年大选宣言中。这一呼吁基于一个简单原则——我们信任新加坡人拥有必要的信息以监督政府。
公民应在《信息自由法》下有权提出请求，并获得公共机构所持有的、请求的详细程度的信息。政府持有的任何能够促进公共辩论的数据或记录，也应在25年后自动解密并向公众开放，当然前提是符合合法的国家安全考虑。
如果没有《信息自由法》和25年后自动解密的框架，政府将不会被迫审查和发布信息，基础性的真相可能会无限期地被掩埋。
事实和解密文件可能被挑选以支持特定叙述。这不是我们希望的新加坡。《信息自由法》和自动解密将把举证责任从公民转移到政府。政府必须证明为何需要保密，而不是公民证明好奇心。这赋予历史学家、记者、公民社会和普通新加坡人监督以他们名义作出的决策的权力。
一个成熟的国家不惧怕自己的档案。我相信工人党提出的建议将加强我们的国家认同，而非削弱。当国家认同建立在事实基础上，即使这些事实复杂或令人不适时，我们才真正建立了国家认同。如果我们真正相信问责治理和知情公民，那么是时候将公众知情权写入法律。
主席：Christopher de Souza先生，请将您的两段发言合并。
历史展览的巨大价值
Christopher de Souza先生（荷兰-武吉知马）：MDDI的核心职责之一是塑造新加坡人的历史叙述，确保我们的国家历史基石准确、有原则且扎实。在这方面，我要对最近关于《信天翁档案》的展览及其背后的团队表示强烈支持，他们值得表扬。
该展览出色地聚焦了独立新加坡诞生的环境、1963年至1965年的时代压力以及我们领导人面临的艰难权衡。
理解这段历史不是学术练习。展览如同指南针，展示了决策如何坚定地基于原则，同时兼具深刻的务实精神。面对生存的不确定性，我们的先驱者坚定不移，做出艰难选择，逆境中建设国家。这些危机中的决策塑造了独立新加坡的基因。
类似的展览还有空间，比如关于COVID-19。毕竟，这场疫情是新加坡独立后历史中的一个近期且决定性的时刻。那些深夜的决策、疫苗采购选择、保护生命和生计，以及我们在2020至2022年间在本议院进行的多次辩论，还有振兴新加坡航空。这是一代人的危机，我们共同克服。
过去危机中的决策塑造了国家的基因，并为新加坡的未来提供了稳定和方向。决策过程应被展示。应通过更多展览汲取经验教训，比如关于COVID-19危机的展览。
人工智能——创新中的辨别力
先生，人工智能具有明显优势。它可以消化并总结大量知识，但归根结底，人工智能是工具。它不能成为主宰。它不会进行道德判断。
因此，在拥抱人工智能时，我们必须保持辨别力。我们应在人工智能促进决策的范围内使用它，但不能让它取代我们的决策。
创新不能以侵犯现有知识产权为代价。在这里，先生，请允许我声明，我是私人律师事务所的合伙人，专注于知识产权法并提供人工智能法律咨询。
规则至关重要。在我看来，必须有明确的界限标志，禁止深度伪造、深度伪造色情、诈骗、欺骗、虚假陈述以及侵犯某些受保护的知识产权。如果新加坡能成为人工智能采用的引擎，同时保持其作为可信知识产权中心的地位，我们就找到了正确的平衡。
简而言之，为了在不削弱人类思维的情况下明智地平衡人工智能的益处，我们必须确保人工智能始终是工具，而非主宰。
人工智能时代重新设计入门级岗位
Abdul Muhaimin Abdul Malik先生（盛港）：主席，政府已投入大量资金支持产业的人工智能转型。在此背景下，我关注当今应届毕业生已受到的影响，以及这对我们国家人才储备的长期复合影响。
在对250家本地雇主的调查中，80%承认人工智能已减少了他们的入门级招聘。鉴于本议院两边议员均已谈及此问题，我不再赘述。
今天，我想提出一项建议，希望政府予以考虑。从企业创新计划到生产力解决方案补助，作为政府支持的交换，必须建立更强的规则，以防止此类支持加速入门级岗位的流失。总体而言，我建议为所有与人工智能相关的岗位和补助计划增加两个条件。
首先，至少应要求公司提交结构化声明，详细说明其人工智能转型预计如何影响人力资源决策，特别是入门级岗位。将淘汰或重新设计多少此类岗位，以及将为受影响员工提供何种职业发展支持？该声明有两个目的——促使公司考虑调整转型以保护自身人才储备，并为政府提供关于入门级岗位在各行业可能受到的定性和定量影响的重要见解。
其次，公司应承诺维持一定数量的入门级岗位，并确保为入门级员工提供结构化发展机会。为便于中小企业合规，第二项条件可仅对超过特定规模的公司强制执行。
为实现此目标，国家人工智能理事会可协调相关机构并咨询业界。例如，MDDI可牵头核实公司是否符合条件。只有获得此核实的公司，方可提交与人工智能支出相关的企业创新计划申报。
先生，我相信我的建议切实可行且必要，以确保公共资金支持的人工智能转型不会以牺牲我们国家人才储备为代价。同样，部分补助计划已有相关条件。生产力解决方案补助要求公司在申请前提交拟议方案的整体影响描述及预期生产力提升。
总之，今天的入门级岗位塑造明日的行业领导者。让我们确保人工智能转型放大而非侵蚀年轻毕业生辛勤争取的职业机会。
降低人工智能和自动化风险
Mark Lee先生（提名议员）：主席，2026年预算正确地将人工智能和自动化置于企业转型的核心。但如果我们希望广泛采用，明确性和商业现实同资金支持一样重要。
许多中小企业并不缺乏雄心，而是缺乏确定性。对于什么算作人工智能或自动化支出；捆绑的数字成本如何处理；机器人硬件和软件层如何分类；以及哪些文件能经得起审计审查，这些都存在不确定性。
当定义模糊时，企业会犹豫。在现金流紧张的环境下，犹豫就变成了不作为。请问数字经济发展局（MDDI）是否愿意与贸易及工业部（MTI）及其他机构合作，确保“人工智能和自动化支出”在操作层面有明确的定义？
如果不明确，我们面临两种风险。第一是“人工智能洗牌”，即标榜为人工智能的支出却没有可衡量的生产力影响。第二是采用不足，企业因合规风险而推迟真正的转型。这两者都会削弱公信力。
然而，仅有明确的定义并不能改变行为。对中小企业来说，问题在于风险不对称。集成、部署、机器人安装和工作流程重设计的成本是即时的，而生产力提升则是渐进且不确定的。
如果我们希望转型超越领先企业，模式必须简单——早期降低风险，强力奖励成果。政府是否考虑加强前期支持，实质性降低早期风险暴露，然后引入绩效挂钩激励，让那些展示持续生产力提升的企业获得更高支持，可能高达合格转型成本的80%至90%？
这并非补贴支出，而是补贴结果——可衡量的每位员工产出、每位员工附加值或成本效率的提升。这样的模式将公共支出与真实生产力提升挂钩，给予中小企业信心去承诺转型。
我们还必须谨慎地塑造国家层面的人工智能话语。公众讨论往往聚焦于生成式人工智能和数字工具，但在劳动密集型行业——物流、餐饮、设施管理、制造业——机器人和先进自动化可能带来更直接的生产力提升。数字经济发展局在塑造这一叙事中扮演关键角色。转型不仅仅是仪表盘和聊天机器人，而是机器人、流程重设计和岗位重设计。
最后，协调至关重要。人工智能相关计划涉及多个机构。从中小企业角度看，现状可能显得支离破碎。数字经济发展局是否考虑加强统一的沟通架构，让企业看到一个连贯的转型路径，而非通过多个机构分散接触？
在结构性紧张的劳动力市场中，生产力关乎生存。因此，转型必须：定义清晰；沟通连贯；激励设计商业合理。如果我们做对了——早期降低风险，果断奖励真实成果——我们就能实现全经济范围的转型，而非孤立的试点流程。
人工智能带来的增长机遇
朱佩玲博士（蔡厝港选区）：主席先生，在我与中风康复者的工作中，我们利用脑成像和机器学习来理解脑损伤后的重组过程。一次扫描可以产生数千张图像。算法帮助我们发现否则会错过的模式。但没有负责任的科学家会盲目依赖模型。我们会严格验证，测试偏差，检查模型失效的情况，因为错误的结论不会仅停留在期刊上，它会影响到个人。
随着新加坡加快人工智能的雄心，我们应将同样的严谨态度应用于国家层面的部署。2026年预算设定了明确方向：由总理担任主席的国家人工智能理事会和国家人工智能任务组，推动先进制造、连接性、金融和医疗保健领域的实际成果。这是正确的姿态：人工智能——不仅是流行词，而是经济战略。
要让人工智能转化为新加坡人切实感受到的增长，有三项纪律至关重要。
第一，价值捕获，而非仅仅采用。预算措施，如扩大企业创新计划以支持合格的人工智能支出，可以促进采用。但采用不等于影响。从“尝试工具”到“重设计工作”的跃迁才是生产力的胜利所在。我们应帮助企业，尤其是中小企业，跨越这一鸿沟，提供与任务相关的行业操作手册、参考工作流程和实用基准。人工智能必须创造企业价值，而非仅仅是技术活动。
第二，信任架构作为竞争优势。在一个碎片化的世界中，新加坡的品牌是严肃系统可靠运行的地方。随着人工智能系统从辅助决策转向塑造结果，保障不能是非正式的。对于高影响部署，应制度化测试、必要时的可解释性和适当的独立审查，以增强信心而不扼杀创新。信任不是创新的副产品，而是我们有意构建的资产。
下午4时
第三，大规模双语人才。我们需要的不仅是人工智能工程师，还需要既懂领域又懂数据的专业人士，理解背景、模型局限和风险。未来的劳动力必须精通代码和背景。如果我们构建价值、信任和双语人才，新加坡不仅会采用人工智能，还将塑造其部署方式，确保我们的增长具有韧性和包容性。我欢迎部长就数字经济发展局如何通过国家人工智能理事会和人工智能任务组推动这些纪律发表看法。
主席：李嘉珊女士，请将您的两段发言合并。
为青年专业人士打造人工智能准备好的中小企业
李嘉珊女士（西海岸-裕廊西选区）：主席先生，这场人工智能转型必须谨慎管理，因为它给许多人带来焦虑。
在我与青年交流中，反复出现两个担忧。第一，被人工智能取代。第二，人工智能采用的适度规模。
关于被取代，年轻专业人士和准备进入职场的青年担心他们的工作会被人工智能取代。我与青年交流时，他们表示担忧无法跟上人工智能发展的速度。随着人工智能使某些任务变得多余，他们的工作也可能被取代。
这适用于准备进入职场的青年和已经在职场的人员。尽管他们是数字原住民，甚至是技术专业人士，但他们仍然紧张，担心难以领先于人工智能的发展曲线。
许多在职人员愿意提升技能，但时间紧张。他们需要灵活的培训、雇主支持和认可，以及每门课程的明确成果。雇主则需要信心和保证，培训能转化为生产力。
所以，问题不在于培训什么，而在于如何让培训切实有效。请问部委有哪些计划，装备劳动力以获得利用人工智能的信心和技能？有哪些计划确保劳动力培训与生产力和业务成果紧密挂钩？有哪些计划鼓励雇主更多支持员工培训？我支持政府探索如何扩大TeSA计划，帮助所有年轻新加坡工人保持竞争力。
关于适度采用人工智能，人工智能的采用不会一刀切。不同企业和行业面临不同限制，尤其是我们的中小企业。众所周知，中小企业雇佣了大多数劳动力，近一半在小型和微型企业。我们不能让他们掉队。但许多中小企业面临真实限制：现金流、回报不确定、人力和实施能力。
因此，支持必须适度且实用。不仅是资金，还包括端到端支持，帮助企业采用、集成、扩大人工智能在核心流程中的应用，重设计遗留系统，并针对不同业务需求定制解决方案。这些人工智能采用的不均衡加剧了青年对工作安全和职业发展的不确定感。
特别是，我想问：部委将如何减少中小企业采用人工智能的不确定性？例如，部委是否会促进与行业协会和高等院校联合开发的共享解决方案或成熟案例的提供？
新宣布的人工智能冠军计划将在支持人工智能融入业务流程方面发挥一定作用。请问人工智能冠军计划将如何与企业创新计划和生产力解决方案补助金等计划协调，考虑到它们由不同部委下属的不同法定机构管理？
图书馆的新可能性
我了解到数字经济发展局正在考虑为我们的图书馆带来新的可能性。我想请求部委在更新图书馆时，将家庭置于首位。先生，我注意到我的发言时间已到。
主席：谢谢。郭振辉先生，您可以将您的两段发言合并。
推动以人工智能为中心的信息技术发展
郭贤川先生（芽笼巴鲁）：主席先生，以人工智能为核心的信息技术发展已不再是理论。在硅谷，领先的人工智能公司和超大规模云服务商已经超越了传统的软件开发。现在的框架包括基于意图的工程、氛围编码和代理式开发。越来越多顶尖程序员公开表示他们不再采用传统编码方式。以人工智能为核心的开发本质上是不同的，它贯穿于设计、编码、测试和持续改进的全过程。
新加坡的信息技术服务公司是我们国家人工智能战略与现实成果之间的交付层。瓶颈不在于人才，我们的技术人员已经准备就绪。问题可能是信息技术服务公司的结构惯性，如果他们不能迅速采用这些新范式，我们的雄心将停留在纸面上。
多媒体发展、设计与创新局（MDDI）能否与已经在本地设立的顶尖人工智能公司和超大规模云服务商合作，将这些技术诀窍转移给本地企业？我们是否应利用这些合作关系，不仅推动企业发展，还要转变我们信息技术公司的软件构建方式？
新加坡政府科技局（GovTech）是否也能迅速采取行动，拥抱这些方法，同时遵守网络安全和监管要求，并逐步要求参与政府工作的IT服务公司也这样做？这并非没有先例。我们曾在建筑采购中强制采用建筑信息模型（BIM），这彻底改变了该行业。
最后，我们如何确保我们的技术劳动力和学生跟上步伐？在技术行业之外，我们如何鼓励跨国公司和中小企业在自身运营中采用人工智能？
支持我们的国家媒体
主席先生，我们的公共服务媒体公司——新传媒、联合早报、CNA、商业时报——不仅是新加坡的真相基础设施，也是我们社会的信任基础设施。在人工智能生成的虚假信息时代，正如一些议员早先提到的，它们是我们人民与被操纵的信息环境之间的屏障。
然而，我们的公共服务媒体面临重大挑战：在碎片化的媒体空间中发行量下降，广告模式快速演变，以及来自海外的虚假信息日益增多。与商业媒体不同，我们的公共服务媒体还承担着国家建设的责任，公平服务所有社区，促进社会凝聚力，维护国家利益。即使是像《华盛顿邮报》这样有声望的国际媒体也不得不采取大幅裁员。
多媒体发展、设计与创新局能否阐述其保持国家媒体吸引力、相关性和繁荣的愿景？它如何帮助媒体保持财务可持续性，以避免类似的痛苦重组？
我们不应低估我们已有的资源。联合早报已是全球最受尊敬的中文媒体之一。CNA的公信力远超本地。通过战略投资，商业时报有望成为东南亚的《金融时报》。我们的公共服务媒体是我们软实力的重要来源。
我们的媒体还必须对所有新加坡人保持相关性，尤其是学生。澳大利亚和英国等国家确保公共服务内容在联网电视平台上保持显著且易于发现。我们能否效仿，使本地优质内容不被算法偏向海外节目所掩盖？
我们的公共服务媒体公司是国家资产。我希望多媒体发展、设计与创新局能分享其保障未来的计划。我注意到我还有18秒，所以想补充一点。我注意到我们要求多媒体发展、设计与创新局在网络安全、国家公共服务媒体、IT服务和人工智能开发方面承担很多责任。我知道预算有限，任务不易完成。提前感谢你们。
主席：陈佩玲女士，请将您的两段发言合并。
数字世界中的信任
陈佩玲女士（海滨组屋-布莱德尔高地）：主席先生，我们是一个开放的社会，无论是物理上还是虚拟上。信息从各个方向涌入，辨别真相从未如此困难。人工智能使这一问题更加严重。
最近，许多人可能读到过网上的假新闻，声称资深部长李显龙公开反对总理黄循财。我一位居民在一月份的区块访问中如此坚信该报道，以至于我很难说服他。此类事件不仅误导信息，还腐蚀相互信任，削弱社会凝聚力，并为诈骗提供温床。人工智能加剧了威胁，因为它可以大规模生成令人信服的内容，快速迭代，并被用来探测和破坏我们的关键信息基础设施。
我们的公共服务媒体在维护事实公共话语中发挥核心作用，是公民获取重要问题真相的首选来源。因此，我有几个问题。
首先，政府将如何支持和加强我们的公共服务媒体，使其能更有效地应对日益嘈杂的信息环境中的假新闻和错误信息？这包括资金、人才培养、编辑独立性以及快速大规模验证的技术能力。
其次，将采取哪些具体措施确保公共服务媒体内容保持高质量和高度可访问性，涵盖语言、平台和人口统计，以便在谣言传播之前，可信信息能传达至每个社区？
同样，我们如何确保真实且经过验证的新加坡叙事传达到国际受众，既保护我们的声誉免受虚假信息侵害，也在全球重大事务中发出我们的声音？
第三，政府是否会为公共服务媒体和公共机构配备先进工具，包括负责任治理的人工智能，用于检测、归因和反制虚假信息？简言之，我们能否用人工智能对抗人工智能，同时设立保障措施避免过度干预、偏见或隐私侵蚀？
最后，除了公共服务媒体，政府将在哪些更广泛的领域投资公共数字素养、快速响应验证实验室以及与平台和民间社会的合作，以增强社会对人工智能驱动的错误信息的韧性？
在技术既带来益处又带来风险的时代，我们必须确保公共信息架构稳健、可信且具适应性。
人工智能治理与主体性
随着新加坡将人工智能视为发展战略必需，我们必须在利用其力量的同时保护人民的长期利益。当前公众讨论多偏向于人工智能“接管”工作和社会的末日论调。这种叙事忽视了一个根本点：我们可以且必须保有人类主体性。我们可以选择人工智能的设计、治理和部署方式。
这一选择需要强有力的国家领导，制定切实可行的治理路径，并持续开展国际合作，明确哪些事项不应委托机器处理。新加坡于2019年1月率先推出了《人工智能治理模型框架》，随后开展了生成式和代理式人工智能的相关工作。这些都是重要基础。
但领导力必须转化为具体行动。包括互操作标准和认证、严格的采购和审计要求、独立监督，以及公众素养和劳动力再培训的投资，使公民能够行使有意义的主体性。在国际上，我们可以推动规范，防止监管套利，确保跨境问责。
因此，在现有框架基础上，政府将采取哪些措施加强新加坡在全球人工智能治理、标准、认证、国际协调和能力建设中的作用，以维护人类主体性，同时负责任地约束人工智能主体性？
主席：陈洁仪女士，请将您的三段发言合并。
数字安全与社会韧性
陈洁仪女士（东海岸）：主席先生，2026年预算强调了一个重要优先事项，即保护新加坡人免受诈骗和网络危害。如今的数字安全不仅仅是避免可疑链接。人工智能改变了诈骗和网络威胁的运作方式。用于欺骗的工具更复杂、更个性化，且更难被发现，即使是那些通常在网络上自信的人也难以识别。
我们现在看到人工智能生成的深度伪造视频，声音可以极其逼真地模仿家人或朋友。高度个性化的诈骗根据个人习惯、脆弱点和网络行为定制信息。错误信息传播速度更快，规模超过事实核查能力。这些风险对老年人、青少年、低收入家庭以及缺乏数字自信辨别真假信息的人群影响尤甚。
为了保障新加坡人的安全，我们必须从“数字安全”转向人工智能风险韧性，赋予人们实用技能、可信工具和强大的社区支持。
我们可以从四个方面加强这方面的工作。我想提出以下建议。
第一，在针对不同年龄组和人生阶段的数字素养项目中，推出国家级的人工智能安全课程。
第二，网络安全委员会（OSC）可以将人工智能特有的风险纳入其在线危害类别，承认由人工智能生成的冒充行为，如深度伪造和大量生产非真实材料，以便受害者能够及时寻求补救。
下午4点15分
第三，成立一个公私合营的人工智能反诈骗工作组，以领先应对不断演变的威胁并协调响应。
第四，利用数字大使的基础，发展一个社区基础的数字安全大使网络，重点关注老年活动中心、学校和社会服务环境中的人工智能风险。
主席先生，数字安全工具旨在保护信任，保护家庭，并确保每位新加坡人，无论年龄或背景如何，都能在人工智能驱动的世界中感到自信和安全。
负责任的人工智能
主席先生，今年的预算正确强调了负责任人工智能的重要性。新加坡已经建立了坚实的基础——从《人工智能治理模型框架》到AI Verify，以及金融和医疗保健领域的行业特定指南。这些是真实的优势，显示了我们对安全和可信创新的承诺。
但随着人工智能成为影响人们生活的日常决策的一部分，新加坡人现在需要更多的清晰度、一致性和问责制。
如今，人工智能已经被用于招聘、信用评估、保险承保甚至公共部门流程。但透明度水平差异很大。许多新加坡人甚至可能不知道何时涉及人工智能。没有定期检查，这些系统可能无意中强化或放大偏见。信任不会自动产生，我们必须有意识地建立它。
同时，我们知道现代人工智能，尤其是前沿模型，复杂且通常是专有的。虽然透明度和独立评估听起来简单，但现实更具挑战性。新加坡现有框架认识到这一点，但部署速度意味着我们需要以实用且适度的方式加强方法。
并非所有人工智能系统都能承担相同的风险水平。一个回答常见问题的聊天机器人与筛选求职者、评估信用或支持医疗决策的算法不同。前沿人工智能模型，最强大且不可预测，完全属于另一类风险。
这就是为什么我认为新加坡应朝着针对高风险或高影响人工智能系统的有针对性要求发展，而非一刀切。透明度并不意味着公开源代码，而是解释系统的功能、风险及所采取的保障措施。只有在潜在危害显著时，才应要求独立审计。
这些并非激进的想法，它们正成为全球规范。欧盟已强制对高风险系统进行审计。加拿大也在朝同一方向发展。美国监管机构要求在金融和医疗领域进行审计。英国正在加强对前沿模型的评估要求。新加坡应保持领先，但方式应符合我们的实际情况并支持创新。
基于风险的方法使我们能够保护新加坡人，同时保持合规的实用性。这确保我们不会给中小企业带来过重负担或减缓创新，同时仍让新加坡人有信心人工智能被负责任地使用。
我的建议是政府与业界共同制定实用指导、沙盒和行业特定标准，建立在我们已有的坚实基础上。这将负责任的人工智能转变为一种共享的国家能力，而不仅仅是监管义务。
企业采用人工智能的准备情况
主席先生，2026年预算为新加坡的人工智能推动注入了真正的动力。但对许多中小企业来说，关键问题仍然存在。人工智能是否能让日常工作更轻松、更高效、更有意义？
中小企业仍面临实际障碍。计算资源成本高昂，数据分散，治理复杂，员工担心工作影响。如果我们不解决这些现实问题，人工智能只会让少数人受益。
更新后的国家人工智能战略为可信人工智能指明了方向，但中小企业需要明天就能使用的实用工具——行业特定的人工智能信任路线图，明确常见风险和良好实践，预先批准的数据处理、模型测试和人工介入检查的治理模板，简单的“绿色通道”指导，使低风险用例能快速推进，而高风险用例获得所需保障。这就是让可信人工智能成为催化剂，而非合规负担的方式。
人工智能冠军计划很有前景，但人工智能采用不仅仅是工具问题，还涉及准备数据、重新设计工作流程和帮助员工建立信心。许多中小企业缺乏这方面的专业知识。
MDDI能否分享中小企业如何利用这些冠军获得治理支持、工作流程重设计，并充分利用国家计算和企业计划？
我们的员工是核心。劳动力转型路线图必须超越广泛技能。员工需要角色特定的技能地图，展示人工智能如何改变任务，明确从当前角色向未来人工智能支持角色转变的路径，以及与中小企业实际采用工具相关的实操培训。当员工看到人工智能如何简化工作并提高生产力时，采用就会变得自然而然。
我欢迎预算对本地人工智能开发者和测试平台的投资，以及政府采购在帮助他们规模化中的作用。随着人工智能越来越多地嵌入运营，新成立的网络韧性中心和加强的中小企业支持将赋予企业安全采用人工智能的信心。
主席先生，当我们结合实用支持、明确的人工智能信任指导、赋能员工、强大的网络安全和充满活力的本地生态系统时，人工智能将成为真正的生产力工具和企业的助推器——也是新加坡的真正竞争优势。
主席：张玉娟部长。
数字发展与信息部长（张玉娟女士）：主席先生，感谢各位议员的发言。请允许我用中文开始回应。
（中文）：[请参阅方言发言。] 主席先生，转眼间，明天就是元宵节了。春节前，我问母亲是否需要我陪她去买新衣服，但83岁的她说：“不用了！我已经在网上找到了喜欢的衣服并下单了。”
我担心她可能会被骗，就问她怎么知道卖家可靠。她自信地回答：“我只有在收到货并满意后才付款。”
在团圆饭那天，她兴奋地给我看她的新衣服，那时我才放心。
主席先生，数字技术给我们的生活带来了许多便利，也为我们的企业创造了新机遇。但它也让我们面临前所未有的风险和危险。同样，人工智能既有利也有弊。几位议员也提到了这一点。
一些新加坡人担心自己跟不上人工智能时代的步伐。我以前也有同感。但正如总理所说，我们不能因为害怕人工智能而停滞不前。
俗话说，“逆水行舟，不进则退。”
其他国家已经发展了他们的人工智能计划。如果我们行动不够快，规划不够广泛，基础不够扎实，我们必然会落后。关键是我们的目标必须明确，措施必须有效。
在这个人工智能时代，我们如何确保新加坡人不被落下，帮助中小企业保持竞争优势？这是我们密切关注的核心问题。
就像我的母亲——她不是数字专家，但在适当帮助下，她也能安全地进行网上购物。
我们不需要强迫自己成为人工智能大师，因为不是每个人都能掌握同样程度的人工智能；他们从中受益的方式也会不同。更重要的是，新加坡必须保持自信，才能在人工智能时代稳步前进。
在今年的COS辩论中，MDDI将提出各种相关举措，确保新加坡人不仅能跟上步伐，还能从中受益。
（英文）：主席先生，人工智能已成为今年预算和COS辩论的焦点。议员们分享了对机遇的乐观和对就业、创造力及自主权影响的担忧。
沙拉尔·塔哈议员提出了一个关于新加坡在人工智能领域独特定位的战略性问题。我们很幸运，国际同行认可我们通过一系列推动因素——从研发和基础设施到安全和治理——在各行业、企业和劳动力中进行整体响应的能力。
在全球舞台上，新加坡经常参与重要讨论。我们对人工智能采取的进步且深思熟虑的态度，使我们成为可信赖的合作伙伴和有价值的参考点。这使得我们能够设定更高的目标。
总理黄循财、副总理颜金勇以及贸易与工业部的同事们阐述了培养人工智能冠军和推进国家人工智能任务的计划。随后，人力部和教育部的同事们将讨论如何赋能当前及未来的劳动力，充分利用人工智能。我将重点谈谈这对我们更广泛的企业基础意味着什么。
简而言之，我们希望充分利用人工智能的普及能力，或者更简单地说，让其利益广泛传播，因为曾经过于昂贵或复杂的解决方案现在变得更易获得。
但如果人工智能沿袭以往技术浪潮的路径，只有少数处于前沿的公司会领先并拉开与其他企业的差距。规模较小且资源较少的企业则需要更长时间才能赶上。然而，它们合计雇佣了我们大多数的劳动力。当它们落后时，风险不仅仅是国内生产总值。风险还包括创业者的希望与梦想、工人的生计以及社区的发展。
这就是为什么数字经济发展局（MDDI）正在创建国家人工智能影响计划——将人工智能的可能性变为多数人的现实，而非少数人的特权。
目前，约有15%的中小企业和大约七成的工人在某种程度上使用人工智能。我们希望鼓励尚未起步的企业迈出第一步，并帮助已经使用人工智能的企业超越基础应用。
在未来三年内，国家人工智能影响计划旨在支持1万家本地企业将人工智能整合到其业务流程中。这将创造一个庞大的早期采用者群体。他们可以成为社区中的倍增者，通过戴妮丝·傅女士向总理提及的中介机构分享经验和知识。
小企业将受益最大。以宏茂桥的一家单店销售商榴莲记忆为例。他们没有条件专门安排团队成员处理客户咨询。结果，饥饿的榴莲爱好者未能得到及时回应，导致销售流失。
但榴莲记忆通过实施一套具备人工智能功能的客户关系管理系统，配备自动回答客户查询的聊天机器人，成功应对了这一挑战。结果，销售高峰期增长了30%。
现在有许多人工智能工具以简单有效的方式改善业务运营。它们占信息通信媒体发展局（IMDA）“中小企业数字化”平台上数字解决方案的30%。我们将通过补助支持扩大人工智能解决方案的范围，以满足不同的业务需求。更多中小企业将能够访问这些预先批准、成本效益高且市场验证的工具，轻松且经济地整合人工智能。
正如普里塔姆·辛格先生和穆海敏先生所期望的，我们希望这些解决方案既具有变革性，又以人为本。同时，李显龙先生担忧人工智能漂绿问题。我们将在补助和激励措施中设置保障措施，同时努力避免规则过于繁重。
一些企业已准备好利用人工智能做更多事情。以Mocha Chai实验室为例。他们是一支才华横溢的多媒体创作团队，提升电影视觉和音效。大多数人不知道，电影中的音效仍然是手工添加的，通常需要四到八周时间。加入IMDA的数字领袖计划并提升技术能力后，Mocha Chai开发了一款新的生成式人工智能工具，能够分析视频素材并自动生成匹配的音效，将数周的工作缩短到仅一天。
这一创新不仅帮助公司节省了成本，还创造了潜在的新收入来源。它为企业及其员工开辟了新的机会。
我们希望有更多像Mocha Chai这样的成功故事。但正如陈洁仪女士、李显龙先生和莎拉尔·塔哈先生指出，更复杂的人工智能应用需要多种因素共同成功。技术往往已准备好，但人们尚未准备好。这就是为什么我们正在加强数字领袖计划，并推出新的数字领袖加速训练营，培养变革管理的技能和信心，而不仅仅是技术能力。
我们也感谢安德烈·罗先生、陈德成先生、法兹利·法兹维先生以及莎拉尔·塔哈先生认可政府已提前规划，管理人工智能广泛使用带来的能源影响。
我们通过多种方式做到这一点。我们在扩展数字基础设施时保持审慎。分配新的数据中心时，我们评估其使用低碳能源的效率。我们正在引入新的可持续性要求，以提升旧数据中心的能源效率。通过国家人工智能研发计划，我们将支持公共研究，推动资源高效的人工智能，深入了解我们的选择。
下午4点30分
随着更多企业采用人工智能，也有机会提升劳动力技能，帮助他们保持竞争力，无论是在入门阶段还是职业后期。除了总理的承诺和人力部的计划外，我想向阿卜杜勒·穆海敏先生、李嘉珊女士和朱佩玲博士等议员保证，数字经济发展局正专注于此。
我们知道专业人士和知识工作者感受到的压力更为明显。但许多人已经找到借助人工智能提高效率的方法。以审计专业人士劳洁琳为例，她在毕马威工作了27年。每次审计，她都要仔细审阅大量文件以评估风险。通过雇主提供的培训，她创建了一个人工智能代理，自动整合公司公告中的关键信息，用于审计审查。
人工智能代理比劳洁琳更快地组织信息，但她的领域知识是确保其查找正确内容的关键。节省了数小时的手工工作后，她现在可以专注于更深入的风险评估，并运用她的人类能力——智慧、校准和专业判断——处理更复杂的工作。
劳洁琳和许多专业人士展示了人工智能技能、领域专长和人类触感的强大组合。我们不可能人人成为人工智能工程师，但我们可以在各自专业领域成为“人工智能双语者”，用人工智能解决领域内的问题。
政府将支持10万名工人成为人工智能双语者。他们将成为有意义的人工智能技能提升的先行者，供他人效仿。我们最初将重点关注高度暴露于人工智能且服务多个行业的职业。信息通信媒体发展局将与相关机构和专业团体合作，扩大其TeSA计划，培养关键领域的人工智能双语工人。我们将从会计和法律职业开始，逐步扩展到人力资源等其他领域。
正如郭振辉先生指出，人工智能也在改变科技行业——许多人现在可以借助人工智能编写代码和构建原型。因此，我们将加强TeSA项目，帮助科技工作者提升价值链，从编写代码到协调由人工智能代理驱动的端到端系统。
随着人工智能快速发展，我们的治理也必须跟上步伐。我们赞同陈洁仪女士和朱佩玲博士关于基于风险、务实的人工智能治理观点。正如苏绍辉先生所言，我们认为人工智能不应取代有洞察力的人类思维。
我们新的代理型人工智能模型治理框架将帮助组织管理能够更独立行动的系统，同时确保有人类监督。我们是全球首个推出此类指导方针的政府。对于高风险、高影响的系统，如前沿模型，我们将逐步加强保障措施。
然而，正如陈佩玲女士指出，本地所做的还不够。最先进的人工智能模型仅在少数几个国家开发，但它们在人工智能安全方面的合作并不深入。
近年来，新加坡举办了多场重要的人工智能会议，促进国际合作。去年，我们组织了新加坡人工智能会议：国际科学交流人工智能安全。此次交流汇聚了来自研究、政府和民间社会的世界级思想家，促成了关于全球人工智能安全研究优先事项的新加坡共识。
最近，在印度人工智能影响峰会上，我分享了新加坡将主办第二届国际科学交流会议，以更新新加坡共识。尽管面临挑战，我们将继续为国际人工智能安全话语做出有意义贡献。
接下来谈谈网络安全。议员们理所当然地关心我们的关键基础设施是否足够防范恶意威胁行为者，尤其是国家支持的攻击者。我想向莎拉尔·塔哈先生和严建业先生保证，网络安全局（CSA）与国内外合作伙伴紧密协作，检测并遏制网络威胁。
在外交方面，新加坡最近圆满完成了第二届联合国信息通信技术安全开放式工作组的主席国职责。
现实情况是，国家支持的威胁行为者司空见惯。然而，达成国际共识，明确什么构成负责任的国家网络行为，仍然非常重要。我们不能指望这些努力能够替代更强大的网络防御能力。在这方面，网络安全局将重点关注三个关键领域。
首先，我们将审查关键基础设施（CII）所有者的网络安全标准和要求。其次，我们将为关键基础设施所有者提供先进的工具，使他们能够应对高级威胁。第三，我们将与合作伙伴共同提升我们网络安全人才的能力。高级国务部长陈杰厚将进一步介绍这些工作。
我们面临的另一个风险是由人工智能等技术助长的虚假信息和错误信息的传播。作为一个多元化社会，我们特别容易受到网络谣言的影响，这些谣言侵蚀了社会和机构的信任。幸运的是，我们一直在加强图书馆和档案馆的建设。它们通过培养阅读习惯和信息素养，帮助培育有辨别力的民众。国务部长拉哈尤稍后将分享更多内容。
我们的公共服务媒体机构在维护我们信息空间的信任方面也非常重要。感谢郭振辉先生和陈佩玲女士对此的认可。我们的公共服务媒体机构覆盖了超过90%的新加坡人。它们仍然受到公众的高度信任，甚至超过了知名的国际和在线媒体。
因此，我们的公共服务媒体机构在对抗虚假信息方面变得不可或缺。MDDI 将继续与我们的公共服务媒体机构紧密合作，以保持其影响力并加强其事实核查能力。例如，CNA 将成立一个数字验证团队。政府机构也与《海峡时报》合作开展了 AskST 系列，以应对虚假信息。
郭振辉先生询问了有关帮助公共服务媒体保持相关性、易于发现以及财务可持续性的努力，因为观众注意力和广告收入正向数字平台转移。
除了提供及时且可信的新闻外，我们的公共服务媒体机构还制作增强我们作为一个民族认同感的内容。它们还通过定期的学生刊物和学校竞赛，在培养青少年的新闻素养方面发挥作用。
鉴于公共服务媒体的重要作用，MDDI 将支持保持公共服务媒体内容可见且易于发现的努力。我们正在研究其他国家的做法，并将与业界协商，确保各项举措合理有效地实施。政府将继续投资我们的公共服务媒体实体，帮助它们在媒体环境演变中发展新能力。
先生，最后总结一下，我们今天所做的投资将决定我们明天是领先还是落后。通过加快人工智能的应用，加强技术治理，并提升人民的辨识能力，我们正使新加坡人能够抓住机遇，共同进步。
主席：高级国务部长陈杰豪。
数字发展与信息高级国务部长（陈杰豪先生）：先生，过去十年里，我们采取了重大举措加强网络安全，例如成立网络安全局（CSA）并引入《网络安全法》以保护我们的关键基础信息设施。
但绝不能自满。我同意维克拉姆·奈尔先生对内政部的看法，即威胁行为者，尤其是高级持续性威胁（APT），只会变得更加复杂。沙拉尔·塔哈先生询问了政府保护我们关键基础设施（CII）的计划。
网络安全是集体的努力。关键基础设施（CII）所有者必须对他们拥有和运营的系统负责。政府也将尽我们的责任。
在本次COS上，我将谈论MDDI的计划，首先，更新网络安全标准和义务；其次，提高关键基础设施（CII）所有者的水平；第三，加强我们网络安全劳动力的能力。
今天，我们对关键基础设施（CII）所有者提出了更高的标准，并对其关键系统或CII系统施加了严格的义务。这是一种经过权衡的做法，旨在平衡国家安全需求与企业成本。我们观察到，威胁行为者也在针对非CII系统，因为这些系统可能安全性较低，且可能成为进入CII系统的入口点。
因此，CSA正在审查当前网络安全标准和义务的范围，可能会包括非关键基础设施（非CII）系统，例如与关键基础设施系统互联的网络。我们注意避免给关键基础设施所有者施加不必要的成本，并将继续采取基于风险、经过校准且务实的方法。
各行业负责人可根据其行业特点引入额外的行业特定义务。例如，鉴于近期频繁发生的攻击事件，信息通信媒体发展局（IMDA）将加强对电信运营商的网络安全监管。IMDA计划就基础设施虚拟化管理和凭证管理等领域提供指导。
我们期望关键基础设施（CII）所有者遵守这些要求。关键基础设施所有者目前会聘请第三方进行审计和定期渗透测试，以验证其防御的稳健性。这些报告随后提交给网络安全局（CSA）进行审查。
除了依赖此类第三方报告外，CSA还希望确保关键基础设施（CII）所有者实施的安全控制不仅在审计期间得到测试和验证，而且能够持续加强。实现这一目标的一种方式是与CII所有者合作进行现场审查。CSA目前正在与各行业负责人讨论实施计划。准备就绪后，我们将联系已确定的CII所有者。
先生，法规和合规只能起到有限的作用。我们需要我们的行业和关键基础设施业主每天持续地尽其所能，保护他们的系统。
在过去的一年里，我走访了关键基础设施（CII）各个行业，花时间与行业负责人和关键基础设施所有者进行了交谈。我们进行了闭门、坦诚的讨论。我们的行业负责人和关键基础设施所有者都明白威胁环境已经演变，并且认识到所面临的风险。然而，他们告诉我，大多数关键基础设施所有者是私营公司，其业务是提供基本服务。他们并非网络安全专家。然而，他们面对的是最顶尖的、由国家支持的网络威胁行为者。一位首席信息安全官告诉我，这就像他带着刀去打枪战。我理解他的观点。
正如我所说，网络安全是集体的努力。我们是一支队伍。因此，政府将积极协助关键基础设施（CII）所有者加强防御能力，更好地应对事件。
通常，国家安全是政府的专属领域，例如开发尖端技术系统和培训熟练操作员以应对各种威胁情景。我们决定将政府的一些专业知识提供给私营部门，以平衡防御者和攻击者之间的竞争环境。我们将帮助我们的关键基础设施（CII）所有者“提升水平”，在与高级持续性威胁（APT）的斗争中自保。
首先是情报。我们将有选择地与我们的关键基础设施业主共享机密威胁情报，以便他们能够更好地发现并迅速应对攻击其系统的威胁。
第二是工具。我们将为关键基础设施（CII）所有者配备专有的威胁检测系统，以增强他们检测网络中恶意活动的能力，特别是针对国家支持的高级持续性威胁（APT）。这些专有工具是对我们关键基础设施所有者目前使用的商业威胁检测系统的补充。我们已经开始在部分关键基础设施所有者中部署这些工具，并将逐步推广到其他所有者。关键基础设施所有者可能需要承担将这些工具集成到其系统中的费用。如有需要，我们将考虑提供资金支持。
下午4点45分
即使采取了这些措施，我们也必须做好准备，某些威胁仍可能未被发现。这就是为什么防御者必须保持警惕，并不断提升他们的能力。
这让我谈到下一个关于创新的观点。威胁行为者也在不断进步。正如沙拉尔·塔哈先生指出的，自治人工智能代理正成为新兴威胁。我们必须同样利用技术来保护我们的关键系统。网络安全局将与关键基础设施业主合作，测试使用人工智能等技术，以帮助提升其网络安全运营的效率和效果。我们将在适当时候分享更多细节。
防御者还需要能够有效地使用这些工具。因此，CSA将与培训机构合作，设计和策划课程，装备网络安全专业人员具备应对APT威胁的专业知识和技能。
保障我们关键基础设施（CII）系统安全的责任不能仅仅落在我们的前线网络安全防御者肩上。这不仅仅是一个技术问题。关键基础设施所有者的董事会和管理层也必须尽其职责。这是一项领导责任。我们将为他们提供相关知识。
自2021年以来，网络安全局（CSA）与新加坡管理大学合作，为高管领导人举办网络安全战略领导力课程。迄今为止，该课程已培训了74名高级领导人，如SMRT的Dewi Anggraini女士、施耐德电气的Andre Shori先生和星展银行的Kang Seng Wei先生。鉴于参与者的积极反馈，CSA将在未来几年内举办更多期领导力课程。我们计划在今年下半年迎来下一批网络安全领导者。
现在让我谈谈我们如何保护公民。就在去年，议员们可能看到过报道，称攻击者未经授权访问了全球数千台物联网（IoT）设备，包括路由器。新加坡也未能幸免。去年，攻击者感染了2700多台设备，如婴儿监视器和路由器。当这些个人设备被黑客入侵时，公民的隐私可能受到侵犯，日常活动也会被干扰。这些设备还可能被不知情地用来发动针对他人的攻击。
政府将采取更多措施保护公民免受这些恶意行为者的侵害。首先，我们将更加确保在新加坡销售的数字产品具备基本的安全保障措施。这将使这些产品更难被攻破。
目前，我们要求家庭路由器必须符合最低网络安全要求。这是因为它们是网络的入口，传输敏感信息。目前要求其符合网络安全标签计划（Cyber Labelling Scheme，CLS）一级标准。CLS类似于家用电器上的能效标签，但它显示的不是能耗，而是设备的网络安全等级。
CLS分为一级到四级，一级为最基本标准。我们已见到威胁行为者使用更先进的技术来利用家庭路由器。因此，CSA和信息通信媒体发展局（IMDA）计划将新加坡销售的所有路由器的最低网络安全要求提升至相当于CLS二级。
除了路由器，网络摄像头也是网络威胁行为者的常见目标。威胁行为者利用这些摄像头监视个人。被利用的图像甚至被上传至色情网站或用于勒索个人。为了更好地保护公民，CSA将探讨要求网络摄像头达到CLS二级标准，类似于家庭路由器。
CSA将继续监测并审查是否应要求更多数字设备达到最低网络安全标准。
其次，对于处理敏感数据（包括个人身份信息）的组织，我们正在考虑引入更严格的网络安全和数据保护义务。政府将带头实施。GovTech将要求管理关键系统和敏感政府数据的政府供应商符合网络信任标志（Cyber Trust Mark）要求。
CSA还将要求以下三类实体符合网络信任标志要求：关键基础设施所有者、对关键基础设施系统进行网络安全审计的审计员，以及CSA许可的提供渗透测试和托管安全运营中心服务的网络安全服务提供商。
相关利益相关者的咨询工作正在进行中，这些措施将在未来两年内逐步实施。
我们也在着眼未来，准备应对明天的威胁。议员Kenneth Tiong希望澄清新加坡对量子安全迁移的态度。我们一直密切关注这一技术趋势。我们也认为后量子密码学（PQC）将成为量子安全迁移的主流解决方案。它经过广泛测试并被国际认可。新加坡将以美国国家标准与技术研究院（NIST）标准为基线。正如Tiong议员指出的，这也是许多其他国家采取的立场。
量子密钥分发（QKD）是一种补充技术，更适合于高保障通信等特定应用。新加坡在量子安全迁移方面采取风险导向的方法。政府正在审查可采取的实际步骤，包括采用PQC及在必要时QKD的适当角色。
我们已开始投资能力建设，以支持企业进行量子安全迁移。2025年10月，CSA发布了《量子安全手册》和《量子准备指数》，以提高对相关风险的认识。我们正与行业专家合作，更好地支持组织的努力，包括通过培训。我们还通过国家量子安全网络加（NQSN+）计划，在全国部署两条量子安全网络。这为企业将量子安全解决方案（如PQC和QKD）集成到其网络和系统中提供了更多选择。
通过支持NQSN+基础设施和服务的提供，我们旨在降低组织实施量子安全解决方案的技术和财务门槛。量子相关技术是一个不断发展的领域。我们将密切关注发展动态，并适时发布相关指导。
如果不同的技术解决方案被证明有效且能满足我们的需求，我们准备予以采用。
先生，我们的数字基础设施支撑着我们的经济和公民的日常生活。数字发展与信息部（MDDI）致力于提升我们数字基础设施的韧性和安全性。
主席：数字发展与信息部国务部长刘洁茵女士。
数字发展与信息部国务部长（刘洁茵女士）：主席先生，在当今世界，许多企业提供无缝、高效且可靠的数字服务。新加坡人也期望政府能做到同样。
虽然我们已取得良好进展，但我们可以做得更好。我们必须承认，公民仍然遇到服务速度不够快、表格要求提供我们已掌握的信息，或系统之间难以互通的情况。我的发言是关于必须改变的内容。
让我借用一个熟悉的童年玩具——乐高积木（LEGO）套装的比喻。我们可以从乐高积木中学到几点。
首先，它们基于对客户及其偏好随时间变化的深刻理解。孩子们成长，注意力持续时间变化，乐高设计也相应调整。
其次，模块化。每块积木设计成能无缝连接。搭建者无需每次都重新发明基本结构。他们可以重复使用、重新组合并向上构建。
第三，客户可以选择构建模型的简单或复杂程度。对一些人来说，简单的乐高DUPLO或乐高城市套装就足够了；而对想要复杂性的用户，乐高Technic或乐高教育SPIKE套装提供了高级机械组件。这些是可选的，但易于添加。
这个比喻听起来简单，但背后的工程学并不简单。对我们公共服务来说，这些教训意味着什么？
我们必须持续不断地了解公民，而非偶尔为之。新加坡人的期望在演变，他们的人生旅程也在变化。十年前感觉直观的服务，今天可能显得缓慢或碎片化。我们必须设计模块化系统和数字组件，使其跨部门协作。当系统无缝连接时，公民感受到的是整体政府。我们还必须能够提供满足复杂需求的服务。我们的先进共享工具，如人工智能平台和编码助手，支持我们大规模构建专业组件。并非所有问题都需要复杂解决方案，但当需要复杂性时，能力必须已具备。安全、集成且准备就绪。
要做好这一切，我们的一些习惯必须改变。有时，数字转型变成了一系列项目。新应用和新试点。转型不在于启动了多少数字项目，而在于公民是否觉得与政府打交道更简单、更快、更清晰。
这需要在问题定义上保持纪律，也需要在试验上保持纪律。在公共服务中，谨慎是自然的。但当期望超前于我们时，什么都不做是有风险的。我们必须学会管理这种风险，而不是消除试验。
通过开放政府产品（OGP）的“为公益而黑客”活动等举措，我们的官员与用户紧密合作，了解并解决实际痛点。一个团队观察到，医务社会工作者在经历情感上困难的谈话后，需要花费大量时间撰写个案记录。他们的第一反应是开发一个自动转录和摘要工具。但该工具并未完全满足用户需求。我们的社会工作者希望能更好地控制个案记录的结构，以便重要信息能够被轻松检索。当生成的记录组织不清晰时，他们会重新编写。
团队将该工具改进为Scribe，一款由人工智能驱动的工具，能够根据用户选择的主题和写作风格转录对话并生成摘要。Scribe现已在100多个社会服务机构和所有公共医疗机构中使用。平均每次对话节省36分钟的文档记录时间。这不仅仅是一个数据点，而是将时间还给了关怀。
如果乐高连接件每年都重新设计，没有人能搭建出连贯的作品。在政府中，不兼容的系统也会产生同样的效果。过去，各机构常为不同需求构建独立系统，认为每个需求都是独特的。虽然出于良好意图，但这导致了重复建设和整合难题。当信息无法跨系统共享时，公民会感受到这种碎片化。
因此，我们的方法必须是模块化的。我们必须提供并使用基于共享安全和弹性标准构建的通用数字组件，如安全登录和支付处理。机构不应重建已有的功能，而应重用、重组，专注于其使命的独特之处。
去年，卫生科学局（HSA）加强了对电子烟的执法，需要一个新的运营系统。HSA基于GovTech和OGP的共享工具构建，如Ownself Gather（案件管理系统）和Plumber（允许官员自动化手动任务，如追踪重复违法行为）。通过这种方式，他们的增强版电子烟信息系统仅用三周时间上线。若从零开始构建，可能需要数月。
我们构建得越快，执法越快，公民的保护也越好。
下午5点
有些服务需要简单、可靠的组件，我们必须抵制过度设计的诱惑。但有时，我们确实需要先进工具，帮助我们更快地构建更好的服务。
Henry Kwek先生询问GovTech如何拥抱以人工智能为中心的IT开发，并鼓励供应商采用此类实践。我们提供通用工具，如支持政府开发者和供应商代码补全的商业AI编码助手。我们还为GovTech官员提供无需编写代码即可构建和部署功能原型的AI工具。无论是公共服务部门还是我们的合同供应商，这些工具均遵循安全开发、安全AI使用和数据保护的标准。GovTech也在内部试点智能AI编码工具，计划将这些能力推广至整个政府。
我们的共享工具还帮助各机构构建更具包容性和以用户为中心的服务。正如Sharael Taha先生强调的，基本政府服务应对所有人，包括残障人士，均可访问。
我们最有价值的共享工具之一是Oobee。它主动检测政府网站的无障碍问题，并建议修复措施，如添加可由辅助技术朗读的描述性文本，帮助视障用户。自2023年以来，Oobee已扫描超过1600个网站。它让我们看到，即使出于良好意图，也可能存在盲点。
我们服务的成功建立在坚实的信任基础上。新加坡人使用我们的数字服务，是因为他们信任这些服务的安全性，并相信自己正在与合法的政府官员打交道。冒充政府官员的诈骗是严重威胁，因为它直接攻击这种信任。
Sharael Taha先生询问我们保护公民免受诈骗的努力。我们已采取措施，例如统一政府短信的发件人ID为“gov.sg”。为方便公民识别并信任政府来电，OGP和信息通信媒体发展局（IMDA）正在开发系统，使各机构可使用以统一前缀开头的号码拨打电话。后续还将显示可识别的来电名称。
许多诈骗者还使用本地SIM卡进行非法活动。为此，IMDA在咨询新加坡警察部队后，最近对所有电信运营商实施了每人最多持有10张后付费SIM卡的限制。政府还将在严格法律保障下，应用分析技术对SIM卡注册数据进行监测，主动发现并阻断潜在诈骗活动。这些措施重点识别可疑的注册模式。
想象一下，从搭建一个初学者乐高套装（可能是一个简单的汽车，几十块积木，设计给儿童）到搭建完整的乐高教育SPIKE机器人。这不仅仅是积木数量的增加，而是技能、自信和雄心的不同层次。搭建者必须随着挑战成长。我们的公务员也必须随挑战成长。
许多忠诚且勤奋的公务员多年积累了工作技能。随着周围工具快速变化，这对一些人来说令人兴奋，对另一些人则感到不安，仿佛他们辛苦培养的专业知识可能在尚未充分发挥前就被取代。
我们的工作不仅是提供安慰，更是提升能力。让他们有信心使用新工具，思考：我能用这个工作，我能提出正确的问题，我能判断输出是否良好。
正如陈振声部长早前提到的，我们已开始为内阁部长和高级公务员领导进行数字培训。领导者设定条件。当他们理解数字环境时，能自信引导变革，向团队提出正确问题。对更广泛的公共服务部门，目标是确保没有公务员在数字世界中感到无力。
数字政府部（MDDI）将与公务员学院共同成立数字政府学院。该学院将为公务员提供数字、数据、设计和人工智能技能培训。我们不仅关注技术，更关注设计以公民为中心且安全的解决方案。
我们还需解决不再支持需求的过时系统。这些系统是在数字化初期构建，采用当时流行的技术。如今，这些系统缺乏灵活性，成本高昂且难以整合，阻碍政策变革，妨碍信息共享，影响无缝服务。我们已开始着手改造。重建系统需要时间，但我们致力于此，因为这是数字转型的基础。每次现代化努力都是重建更快、更互联系统的机会，更好地支持服务交付。
我已描述了提升政府数字服务的必要措施。我们必须持续了解公民，采用模块化构建，并培养应对复杂性的能力。我们将提升公务员技能，并逐步重建许多过时系统。
如果我们做得好，公民将感受到不同。他们与政府的互动将更简单、更清晰、更有人情味，尤其是在公民精力有限时刻。
以新生儿父母为例。在那宝贵的最初几天，最想要的是与宝宝共度时光，而非填写多份表格。因此，我们围绕这一生命时刻整合了服务。通过LifeSG，父母可以完成出生登记、申请婴儿奖金和共享育儿假，减少大量文书工作。目标简单：步骤更少，重复更少，家庭时间更多。
借助人工智能，我们可以更进一步，从被动响应服务转向主动引导服务。
以SupportGoWhere门户为例，它整合了31个机构的政府计划。想象一下，面对近300个选项，涵盖不同生命阶段和需求，寻找支持的难度。我们的长者及其照护者告诉我们，他们感到信息过载，路径繁多，不知从何开始。因此，我们重新设计了体验。长者支持推荐器会询问他们几个问题，然后利用人工智能推荐最相关的计划。我们仍在完善中，但这正是我们的方向：打造一个帮助公民找到所需信息的政府，而非让他们四处寻找。
提供更优质的服务也意味着我们宝贵的公共服务人员可以将时间和资源集中用于支持那些确实无法使用技术的人群。他们可以更及时地得到我们工作人员的服务。
更好的服务还必须更快捷，因为延误不仅仅是行政上的问题。它们确实影响着真实的生活。这一点在医疗保健领域尤为重要。
随着我们人口老龄化，对医疗专业人员的需求将增加。繁琐的人工注册流程可能成为引进护士的瓶颈，延误护理。因此，我们正在简化流程，重建专业注册系统，以简化和自动化对医疗专业人员的常规审核。我们已将外国护士注册的处理时间从最长六个月缩短至30天。对于等待护理的患者和家庭来说，这意味着更早的治疗、更早的支持和减少焦虑。
这些例子不是值得赞赏的例外，而必须成为常态。我们所有政府工作人员都必须自问：如果我们负责一项政策，它是否以最简便的方式实施？如果我们设计一项服务，要问自己：我会接受这种体验给我自己的家人吗？
我们的转型不会一蹴而就，但这是我们将坚持的标准。
主席先生，建设数字能力不仅仅是追逐技术，更是提升我们的服务标准。新加坡人越来越多地将我们与他们日常生活中最好的数字体验进行比较，而不仅仅是与其他政府相比。这是标准，我们有责任达到。
主席：数字发展与信息国务部长拉哈尤·马哈赞。
数字发展与信息国务部长（拉哈尤·马哈赞女士）：主席先生，虽然技术让我们的生活更便捷，并承诺更美好的未来，但我们必须确保我们的数字社会保持安全和充满活力。
政府将继续在保护最易受网络伤害的人群方面发挥重要作用。同时，我们也必须赋能公民，培养他们在当今数字世界中，尤其是在人工智能兴起背景下，导航和学习的技能与信心。我将概述数字发展与信息部在这些领域的努力。
让我先谈谈网络伤害。我们许多人都听说过，甚至认识遭受网络伤害的人。有些受害者经历过网络跟踪，有些人的私密照片被滥用。受害者及其家人常常承受巨大痛苦和无助。
这就是为什么去年十一月议会通过的《网络安全（救济与问责）法案》如此重要。
作为《网络安全（救济与问责）法案》的一部分，我们将成立一个新机构——网络安全委员会（OSC）。OSC将在今年上半年成立。它将首先支持五类高度普遍且严重的网络伤害受害者：网络骚扰、私密影像滥用、基于影像的儿童虐待、网络曝光（doxxing）和网络跟踪。经评估受害者报告后，OSC可发出指令，禁用有害网络内容的访问或限制施害者的网络账户。
即使我们设立OSC提供额外支持渠道，我知道许多家长自然担心孩子的日常数字活动。儿童是最活跃的数字用户之一，许多家长在监控孩子数字使用与其他事务之间疲于奔命。数字发展与信息部2025年的数字育儿研究显示，超过半数受访者希望政府提供更多支持，包括更强有力的立法，帮助他们管理孩子的数字活动。
何亭如女士询问了更好保护儿童免受社交媒体相关伤害和风险的努力。
我们采取了渐进的监管措施，回应家长和社会的关切。过去三年，我们推出了两项《网络安全行为准则》。准则要求指定的社交媒体服务和应用商店尽量减少新加坡用户，尤其是儿童接触有害内容。准则还要求指定社交媒体服务和应用商店向信息通信媒体发展局（IMDA）提交年度网络安全报告。IMDA目前正在评估2025年指定社交媒体服务提交的年度网络安全报告。IMDA的总体报告将在指定社交媒体服务报告准备好后一并发布。
下午5时15分
目前，我们在现实世界中已有年龄验证措施，比如超市或便利店在销售限制年龄的商品（如酒精或烟草）前检查顾客身份证件。从本月底起，指定应用商店必须实施年龄验证措施，防止18岁以下用户访问和下载不适龄应用。
随着新风险不断出现，网络安全仍是全球持续的挑战。一些海外地区已宣布或实施社交媒体禁令。新加坡也希望加强对儿童的网络保护，我们希望正确且全面地推进。
数字发展与信息部继续研究社交媒体禁令的影响，计划将年龄验证要求扩展至指定社交媒体服务。这将更好地确保网络服务对用户，包括儿童，适龄。与指定社交媒体服务的磋商正在进行中，更多细节将于今年晚些时候公布。
政府对应用商店和社交媒体服务之外的网络伤害保持警惕。一些家长表达了对网络视频游戏带来伤害的担忧，包括接触不当内容、网络欺凌和屏幕成瘾。我们认识到这些关切，正在研究是否需要对网络视频游戏实施保护措施。
我们也关注其他可能威胁网络安全的在线服务。正如何亭如女士指出的，一个例子是人工智能被滥用生成不雅内容，如性内容和暴力内容，且实时大规模生成。嵌入社交媒体服务的聊天机器人带来独特风险，用户包括儿童更易访问。
何女士还提到一个令人担忧的趋势，即用户利用X平台的聊天机器人Grok的提示，替换成人和儿童的衣服为暴露服装，如比基尼。信息通信媒体发展局正在与X平台就此问题进行接触。我们注意到X已在全球范围内采取了一些措施应对此事。我们将继续密切监控，并与X合作提升新加坡用户在其平台上的网络安全。若指定社交媒体服务未遵守社交媒体服务行为准则，我们将毫不犹豫地追究责任。我们也在研究是否需要对人工智能聊天机器人实施保护措施，以更好地防范滥用带来的伤害。
虽然家长可以期待更强的保护措施来防止儿童遭受网络伤害，但家长在培养儿童健康数字习惯方面也扮演重要角色。数字时代的育儿无疑充满挑战。我们经常听到孩子在家庭晚餐时沉迷设备，或家长感到被排除在孩子的数字空间之外。有些人甚至形容当今育儿如同逆流而上，在多重优先事项中挣扎求生。这些担忧是真实存在的，我们希望家长知道他们并不孤单。
为应对这些担忧，数字发展与信息部已推出家长资源，并加强努力使其在社区更易获取。家长可通过IMDA的“数字生活”门户获取指导孩子数字互动的建议。这些建议针对育儿旅程中的不同数字里程碑，如孩子首次使用设备、首次使用社交媒体和首次玩网络游戏。
有幼儿的家庭还将获得数字育儿工作坊和网络研讨会的支持。这些课程设计满足不同需求，有的支持幼儿家长，有的则面向可能遇到更复杂网络情况的青少年家庭。我们将继续加大对数字育儿的支持力度，欢迎大家提出改进建议。
为数字世界做好准备，不仅要求我们成为安全的网络用户，还要成为有目的、有辨识力的学习者。正如郭振贤先生和李慧颖女士所强调的，我们需要为学生和教育者准备一个人工智能赋能的未来。教育者在培养学生关键技能方面发挥着关键作用。
国务部长刘洁敏早前谈及提升公共部门数字能力。我们也在加大力度提升教育者对科技的知识和理解。去年，数字发展局（MDDI）与教育部（MOE）推出了智慧国教育者奖学金，首届共有58名学员参加。我很高兴收到的反馈积极。许多参与者表示，智慧国教育者奖学金提升了他们引导学生成为有思考力和责任感的科技使用者的能力。
裕廊中学历史科首席教师易扎尔·萨尼是智慧国教育者奖学金的学员之一。在一项关于新加坡华人移民社区的中一历史探究项目中，易扎尔的学生利用人工智能进行历史“苦力”访谈，并使用ElevenLabs或谷歌NotebookLM制作自己的AI播客。此类举措展示了教育科技在课堂上的强大力量，学生们批判性地评估从AI工具获取的信息，以提升学习效果。
今年，我们正在优化该奖学金，聚焦人工智能的力量与可能性。通过工作坊和行业参访，教育者将更好地理解AI在职场中的相关性，从而支持学生发展AI技能和能力。我们将更新学校课程以满足新兴需求，正如我们以往所做的那样。正如总理所言，AI素养是基本的数字能力，未来将变得更加重要。
沙拉尔·塔哈先生和达里尔·大卫先生询问政府计划如何让每个孩子无论背景如何都能获得数字技能和工具。我们将继续确保AI素养项目在学校中保持可及性。
数字发展局正与教育部合作，更新小学和中学的“Code for Fun”项目，将AI技能纳入所有学生的核心基础能力。该项目将于2027年向所有学校推广。小学生将学习AI基础知识，如制作数字故事书；中学生将学习利用AI为现实问题创造解决方案。在学生尝试和学习AI的过程中，也将了解其风险、局限性及负责任的使用方式。
对于低收入家庭，数字发展局的DigitalAccess@Home计划将继续通过补贴宽带和计算设备给予支持。
感谢法兹利·法兹维先生、陈洁仪女士和沙拉尔·塔哈先生对加强公民数字素养和韧性公共教育举措的关注，包括对长者和残障人士的支持。如今，公民可以通过数字发展局的“数字技能终身学习”资源掌握在数字空间中导航的技能，包括如何自信且安全地使用生成式AI，以及如何识别AI风险，如错误信息、诈骗和深度伪造。
我们的图书馆也是数字学习的重要接触点。国家图书馆局（NLB）的S.U.R.E项目（即来源、理解、研究、评估）鼓励新加坡人评估信息的可信度和可靠性。NLB将推出新的资源包和推广项目，提升信息素养技能。NLB还将在公共图书馆及其他公共场所举办巡回体验展，公众可体验AI的用途和益处，以及如何安全、负责任地使用生成式AI。
我们将继续为弱势群体提供有针对性的支持。在新加坡各地的SG数字社区中心，长者可以学习如何使用数字服务处理日常事务，如预约医疗和手机银行。我想向林淑仪女士保证，政府将继续采取“数字优先，但非仅数字”的策略。需要面对面支持的公民，尤其是长者，仍可在政府机构的实体服务点和ServiceSG中心获得帮助。
我们还将继续与“数字生活”合作伙伴协作，帮助残障人士有意义地参与数字世界。例如，新加坡导盲犬协会开发了一套工具包，帮助视障社区成员学习使用智能手机上的低视力辅助功能，如VoiceOver功能。
通过这些努力，我们正在建设一个包容的新加坡，让每位公民都能从我们的数字未来中受益。先生，请允许我用马来语说几句话。
（马来语）：[请参阅本地语演讲。] 主席先生，掌握新技能有时会感到艰难，但在同伴和社区的支持下，学习体验会更加有意义。数字技术，包括人工智能，已成为我们日常生活和未来工作的关键组成部分。因此，我们每个人不仅要具备数字技能，还要自信且深思熟虑地使用它们。每个人的数字旅程处于不同阶段，有些人刚开始探索和尝试，有些人则专注于提升技术技能并付诸实践。
为了实现我们对AI自信的马来/穆斯林社区的愿景，M³推出了由MENDAKI基金会领导的Langkah Digital计划。正如我上个月宣布的，Langkah Digital围绕三个关键元素设计——Kenal（认识）、Guna（使用）、Yakin（自信）。Kenal帮助人们了解安全的数字探索；Guna鼓励将技术融入日常生活；Yakin培养终身数字学习，使社区能够独立适应。这种循序渐进的方法让我们能够接触社区不同群体，提供符合其数字旅程阶段的支持。
秉持我们的“gotong-royong”（互助）精神，MENDAKI将汇聚整个社区，包括M³家庭、马来/穆斯林组织、MENDAKI专业网络，以及来自公共和私营部门的合作伙伴。一个好例子是AI工程师卢克曼·努尔·哈基姆，他主动支持Langkah Digital。
卢克曼在阿尔凯尔清真寺主持了超过30名参与者的AI课程，并为超过60名参与者举办了AI工作坊。这些项目满足了社区中不同能力和兴趣的个体需求。有些介绍了ChatGPT等AI工具，有些则专注于更高级的技能，如提示工程。
像卢克曼这样的数字冠军是Langkah Digital的推动力。他们不仅帮助不同群体聚集，还促进围绕技术的有意义对话与合作。
尽管该项目刚于上月启动，我很欣慰我们已在社区举办了12场与AI相关的工作坊和活动，参与人数超过400人。我希望Langkah Digital能赋能更多社区成员。
（英语）：主席先生，随着技术不断发展，单纯提升技术技能已不够。要明智地利用AI，我们需要提出正确的问题，并对得到的答案保持辨别力。因此，我们的孩子需要从小培养有效阅读和处理信息的能力。
我强调阅读，是因为在信息以极快速度、常以短格式和视觉形式呈现的世界中，阅读技能正日益面临风险。阅读是学习新技能的基础。它提升注意力，培养批判性思维，激发创造力和同理心。这些都是我们利用技术造福自己和他人的关键素质。因此，推广阅读对于应对议员们提出的AI可能带来的社会和智力退化问题至关重要。
李嘉欣女士询问国家图书馆局如何以家庭为中心刷新图书馆的角色。我们可以从鼓励家长从小培养孩子良好的阅读习惯开始。这也为我们的孩子在设备泛滥的时代提供了无屏幕的替代选择。
这就是为什么国家图书馆局将继续与教育部合作，加强学校图书馆项目。这包括即将举办的未来学校图书管理员峰会和网络研讨会，赋能学生图书管理员成为阅读倡导者和信息素养的推广者。学生图书管理员还可以通过志愿服务、学生实习和学习之旅拓宽学习。国家图书馆局将加大力度培养良好阅读习惯，相关计划将适时公布。
我们的图书馆还在保存和分享新加坡故事方面发挥着重要作用。这些故事构成了我们社区的基石，确保我们的集体经历不会被时间遗忘。
下午5时30分
为纪念新加坡独立60周年，国家图书馆局（NLB）和数字发展局（MDDI）联合推出了《信天翁档案》一书及展览，记录了围绕新加坡独立历程的事件、人物和辩论。我感谢Christopher de Souza先生对参与团队的赞扬。该展览在公众中反响强烈，自去年12月以来，已有超过13万人次参观；96%的参观者表示他们对新加坡走向独立的道路有了更深刻的理解。
针对Fadli Fawzi先生指出展览内容与他在学校所学不同的观察，新加坡与马来西亚分离一事历来存在不同观点。这并不令人意外，鉴于历史记载的性质。
例如，1990年代初期解密的英国、澳大利亚和新西兰档案反映了其外交官和政府的视角。1998年，新加坡官员的完整视角版本也得以公开。Albert Lau教授的《痛苦时刻》仍是关于分离事件最权威的记述。新加坡开国总理李光耀先生的回忆录第一卷也包含了引人入胜的叙述。
正如《信天翁档案：分离内幕》编辑说明中所述，理解分离事件所需的重要材料均未被隐瞒。所公布的文件均为完整复印，未作删节。公众，包括议员，可前往新加坡国家档案馆查阅所有解密材料，自行形成细致入微的观点。
Fadli先生还提出了《信息自由法案》的建议，主张自动解密并在25年后公开发布档案。
一些国家实施类似立法的经验显示，大多数甚至全部都会有例外条款。事实上，实施《信息自由法案》可能无意中导致更多不透明。托尼·布莱尔2010年回忆录中记录了他对英国《信息自由法案》的看法，该法案在他任首相期间颁布。他卸任后表示：“《信息自由法案》……是一部危险的法案，因为政府需要能够在保密环境下辩论、讨论和决定问题。没有保密，人们会有所顾忌，选项的考虑会受到限制，这不利于良好决策。在所有走这条路的体制中，结果是人们会注意自己写下的内容，谈话时不愿写成文字。这是分析复杂问题的极差方式。”
换言之，《信息自由法案》可能阻碍而非促进治理，因为被视为过于敏感的问题根本不会被记录。因此，我们的出发点应是优先实现促进良好治理和知情公民的透明度，而非单纯追求透明。
我们已有机制允许公众申请查阅政府档案以供参考或研究，历史学家和研究人员也利用此机制提名文件进行审查。政府已逐步向公众开放更多档案，并将继续如此。
新加坡故事不仅加深我们对历史的理解，也使我们能够重新想象未来。去年下半年吸引了两百万访客的SG60心灵体验，让新加坡人表达对国家的希望和梦想。我很高兴听到该体验反响良好，大多数访客给予五分满分评价。
该体验也是利用数字创新的范例，如人工智能和沉浸式叙事，呈现多感官体验。结合国家图书馆局的资源，访客有机会学习如何在日常生活中使用人工智能。
最后，主席先生，随着数字技术日益复杂和先进，其改变社会和生活的潜力，无论好坏，前所未有。
正如马来谚语所言：“Berat sama dipikul, ringan sama dijinjing。”我们共同承担责任，携手克服大小挑战。善用技术利好并减轻其负面影响，需的不仅是技术知识。最终，我们需要全社会共同努力，汇聚政府、产业、学术界、公民社会和公民的经验与视角。
让我们携手为子孙后代开创光明而充满希望的数字未来。[掌声]
主席：我们大约有25分钟时间进行澄清发言。我将优先安排已提交发言时间的议员，按其申请时间长短排序。Sharael Taha先生。
Sharael Taha先生：谢谢主席。我感谢部长及政治职务持有人详尽的回应。我有几个补充问题。
我相信部长还未有机会回应我关于国家人工智能理事会（NAIC）角色的发言。NAIC的具体职责是什么？它是否拥有执行权，还是负责跨部委实施监督，或仅为咨询性质？NAIC如何整合贸工部（MTI）下的经济战略与数字发展局（MDDI）下的数字治理，确保协调推进？
另外，关于国家人工智能战略2.0（NAIS 2.0），目前进展如何？我们从实施NAIS 2.0中学到了什么，这些经验如何影响当前计划？
我最后一个补充问题是关于信任和监管公信力。Josephine Teo部长提到新加坡以信任和监管公信力著称。我们如何将这种信任转化为实际操作？是否考虑为金融、医疗和航空等高风险人工智能系统建立正式的可信AI认证制度？这或许能成为我们的差异化优势和竞争力。
主席：Josephine Teo部长。
Josephine Teo女士：主席，问题不少，我会尽力回答。
或许先回应Sharael Taha先生关于我们从NAIS 2.0实施中学到的经验。总理于2023年12月启动了NAIS 2.0，至今刚刚超过两年。如果说到目前为止的学习，或许有两三个关键观察。
首先，或许可以参考贸工部早前发放的资料——关于“人工智能巷”（Kampong AI）的例子。我记得2023年左右，我们曾访问旧金山一个被称为“Cerebral Valley”的角落。那里极具吸引力，因为黑客屋充满活力。大学早晨发布论文，晚上就有讲座，人们会说这是我的应用方案。一周后，原型已建成，风险投资者被邀请展示。你能感受到那种令人羡慕的热潮。
我们问自己：我们能做到吗？当时我们并不清楚本地生态系统是否足够全面，能围绕人工智能活动形成中心。于是我们决定从“人工智能巷”开始。
起初并未称为“人工智能巷”，我们只是决定必须将社区聚集起来。那时我们还没有“Cerebral Valley”，只是想看看能否起步。结果发展速度远超预期，规模也更大。去年，“人工智能巷”举办了约150场活动，吸引约4000人。正因如此，我们有信心推进“人工智能村”。
但反思和经验是什么？就是可以从小做起，但要有远大梦想。不必总是采取一蹴而就的方式。这是第一点观察。
第二点观察是关于我们如何落实想法，实际上，差距是机会。让我稍作解释。
在开展“人工智能巷”后，我们也在建设人工智能卓越中心。如果没有人工智能卓越中心的良好势头，推动人工智能冠军计划将很难起步。没有基础，难以推进。
其中一个卓越中心专注制造业。观察该中心的活动后，我们意识到某些行业卓越中心可以解决共性问题。这使我们想到某些行业已具备端到端转型的准备。
因此，国家人工智能任务也在此基础上发展而来。随后我们自问，产业界已有活动，政府也有人工智能活动，正如国务部长Jasmin Lau所述，研究领域也积极发展。
微软在新加坡已多年，但直到去年才设立研究实体。谷歌DeepMind也决定在此开设办公室。
这告诉我们，研究活动正在加快步伐，因为人工智能顶尖人才被大胆的抱负和愿意解决重大问题的精神所吸引，这些问题困扰着许多人。所以，我们需要有这样的精神。
这些差距，作为机会，特别适用于人才领域。我们已经开始广泛思考人工智能创造者。比如微软研究院、谷歌DeepMind的人才。例如，在谷歌DeepMind的新加坡团队，他们正在为最新的Gemini模型的发展做出贡献。
所以，我们考虑过人工智能创造者，也考虑过人工智能实践者——机器学习工程师、数据科学家——还考虑过用户——广泛的基础。在实施卓越中心的过程中，我们也意识到，在这之间，需要既理解人工智能（不一定是最深层次），又具备良好领域专业知识的人才。有一种观点认为，需要“通双语”的人才——既懂人工智能，又懂其领域知识，并能将两者结合起来。
即使在我们思考人工智能人才光谱时，原本的差距现在反而成为我们可以采取行动的机会。这是第二个反思。
第三点特别针对您关于国家人工智能理事会（NAIC）角色的问题。
有一句广为人知的话，我们常用它来谈论社会团结。这是非洲的一句谚语：“如果你想走得快，就独自走；如果你想走得远，就一起走。”
事实证明，这句话同样适用于人工智能的发展。如果我们只局限于各垂直领域内部的努力——比如MDDI做这个，MTI做那个——我认为我们各自都会有效率，但无法实现协同效应。Lorong AI可以变成Kampong AI，因为有JTC。如果JTC没有抓住这个理念，这种增长就不会出现。
因此，NAIC正是为了做这些事情。把整体聚合起来，使整体大于部分之和。
如今，各机构已经开始合作，但往往是一对一的合作，而我们需要的是多对多的合作，并且需要机制来理清优先事项、资源投放、探索是否走入死胡同，以及可能需要的政策调整。
所以，NAIC不仅仅是一个咨询角色。我们的总理事务繁忙，他在很多方面给我们提供建议，但他担任国家人工智能理事会主席，是一个非常强烈的信号，表明我们热衷于让人工智能转型惠及所有新加坡人，而不仅仅是停留在愿景层面。我们必须将其变为现实。这就是它的目的。
下午5点45分
主席先生，很快回答，关于信任和监管公信力的问题，以及我们是否会设立相关机构。
事实上，我们已经采取了措施。例如，除了引入模型治理框架外，我们还设立了数字信任中心，作为指定的人工智能安全研究所。我们还有在线安全先进技术中心。这些机构在学术组织内发展起来，但我们会寻求与产业界、学术界合作的机会，看看如何将这项能力制度化，并转化为新加坡的优势。
主席：Jessica Tan女士。
Jessica Tan Soon Neo女士：先生，首先，我感谢部长谈及人工智能影响计划和人工智能民主化。我认为这非常重要，所以我想从个人和公民的角度澄清一些问题。普通公民如何期待人工智能影响计划？他们如何参与？他们该如何利用这个计划？
我还有一个相关问题，关于人工智能的数字素养。我提到过生命周期阶段的问题。我们在为公民甚至员工考虑数字素养时，如何确保从生命周期阶段而非仅仅年龄或个人档案的角度来考虑？
还有一个关于负责任人工智能的问题。鉴于人工智能现在被广泛使用，且在许多关键决策领域应用，政府如何确保并将如何确保偏见不会被延续？最后一点是关于包容性。鉴于人工智能将深刻融入我们的生活，无法忽视，我们如何考虑多语言支持，帮助那些英语不太流利的人？
Josephine Teo女士：主席先生，关于公民如何接触国家人工智能影响计划，这不仅仅通过MDDI和IMDA监管的渠道。我们会与行业协会、商会以及专业团体紧密合作进行推广。这些中介机构对此非常感兴趣，我们欢迎他们的参与。
我们已有现有项目作为良好基础。例如，企业解决方案主要通过生产力解决方案补助计划提供。许多企业对该补助计划非常熟悉，中介机构也熟悉。我之前提到，大约30%的解决方案现在具备人工智能功能。由于兴趣和工具的可用性，这一比例可能会增长。
如果我们看已经使用数字工具的企业，议员可以想象供应商会告诉企业，“如果你在此基础上增加这个人工智能工具，实际上你能做更多以前做不到的事情。”这是一种在已有良好基础上推广的方式。
对于个人，我提到我们将从会计和法律专业开始，然后扩展到人力资源等领域。比如人力资源领域有大约5万名从业者。通过他们自己的机制，也能接触到个人。这些是我们推广国家人工智能影响计划访问的不同途径。
我们欢迎建议。如果有人来告诉我们“你们可以这样做”，我们非常愿意了解如何做。我相信我们的高等院校也非常参与，因为他们与私营部门互动，许多学生有兴趣原型设计解决方案并实施，我们也会关注这些。
关于议员提到的负责任人工智能偏见问题，如何确保不被延续。我们对人工智能潜在风险和危害的应对是多管齐下的。在某些情况下，我们已有的保障措施可能需要加强。例如，人工智能可能助长的色情内容、儿童性剥削材料，这些在刑法中已被禁止。我们可以更新刑法以明确规定。我们也会考虑类似《职场公平法》的情况，如果招聘过程中出现偏见，可以依靠该法，因为它不依赖于不公平或歧视产生的具体方式，我们可以这样处理。
但也有现有法律和措施不足以应对的情况。例如，我们引入了《选举（网络广告诚信）法案》（ELIONA），专门应对选举环境中针对候选人的人工智能生成内容。我们将继续保持灵活应对，利用现有法律，必要时加强法律，也不犹豫在明确可行时引入新措施。
这些就是我们的做法。关于包容性，我完全同意您的观点，多重包容性。我也完全同意多语言支持——或许我会请数字发展与信息高级国务部长陈杰豪先生回应，因为他一直在这方面工作。
数字发展与信息高级国务部长（陈杰豪先生）：先生，关于Josephine部长提到的包容性，特别是语言方面，我来详细说明。
众所周知，新加坡是一个多语种社会，不同社区对他们的语言，尤其是母语，感到非常自豪。MDDI工作的一个部分，特别是MDDI中的“I”部分，是关于信息的。我们设有一个委员会，汇集了学者、媒体代表、学校人员，围绕翻译工作。这就是国家翻译委员会，我协助监督该委员会。
我们正在进行的一个项目，是关于翻译的，这也是我们如何思考为人工智能提供实用工具的一个很好的例子。翻译不仅仅是使用任何工具、任何大型语言模型（LLM），因为你需要本地语境、本地细微差别，例如，并非所有人都理解“chope the table”（占座）的意思。
那么，如何将一些本地术语，特别是本地特定语境的术语，翻译成不同的母语语言呢？国家翻译委员会一直与GovTech和科学技术研究局（A*STAR）合作，针对多个大型语言模型进行本地语境的微调。目标之一是希望在今年晚些时候拥有一个模型，使不讲英语的人士能够更包容、更便捷地访问我们的面向公众的网站和其他资料。并支持公共服务机构更轻松地将许多以英语编写的材料翻译成不同的母语语言——马来语、普通话、泰米尔语。
我们如何在整个公共服务系统中大规模地实现这一目标？因此，我们正在开发一个人工智能工具，希望在今年晚些时候分享更多细节。这只是我们使用人工智能技术的一个具体例子。我们不仅仅是“复制粘贴”其他国家的做法，而是针对本地语境进行微调、适应和定制，以创造价值并为所有新加坡人增添包容性元素。我们希望稍后分享更多细节，并希望获得所有议员的支持。
主席：我们快到截止时间了，我再请一位议员提问。黄伟贤先生。
黄伟贤先生：主席，我想问MDDI，鉴于我们正在大幅改变与市民互动的方式，MDDI如何设定并传达正确的期望值。我赞赏MDDI在电子服务转型方面的雄心和决心。简而言之，我们要求政府用更少的资源做更多的事；做得更快；做得不同，不仅仅是与政府交易，还包括向民众提供建议，比如SupportGoWhere。因此，这是一场政府服务方式的重大转型。
而且，我们不是通过编码开发服务，而是要求公务员协调人工智能代理进行编码。这是相当不同的方式。新的做法总是存在初期不顺利的风险。那么，我们如何设定正确的期望？我认为劳明霞国务部长提到，新加坡人对电子服务有很高的期望。
也许我们可以借鉴谷歌的做法，当他们向我们所有人提供服务时，会有谷歌Beta版、谷歌预览版，用这些方式向民众传达某些服务——
主席：黄伟贤先生。
黄伟贤先生：——可能是试验性质，我们如何更好地传达期望？好了，我说到这里。
主席：张燕菁部长，请简明回应。
张燕菁部长：劳明霞国务部长可以补充，但我先简要谈谈市民的期望。我们并不认为这是负面的。议员和我都经常使用数字服务。作为市民，希望政府在数字政府服务交付方面做得更好，是很自然的事情。因此，我们希望对自己有更高的标准。
但议员的观点非常有用，需要牢记的是，关于失败的期望，关于我们可能需要引入更多安全措施而牺牲一些便利性的期望，这是我们需要向更多人普及的。
在网络安全领域，我们常说这是一个权衡。当事情能够非常快速推进时，你会担心安全功能是否已被妥善内置，有时我们需要优先考虑安全而非便利。因此，对于我们可能提供的一些服务，我们并不确定是否需要那种措施，我们会尝试不故意制造不便。但如果确实需要，我们必须向民众解释并妥善处理。我不知道劳明霞国务部长是否想补充。
主席：劳明霞国务部长，您有一分钟时间。
劳明霞国务部长：感谢黄伟贤先生的提问，这让我们有机会向新加坡其他市民解释我们如何努力优化。我刚才提到更快更好的服务，但非常重要的是，我们不能只把数字转型看作一两个指标。我们的系统必须既快速又安全，高效又有韧性，雄心勃勃又值得信赖。因此，我们不能只优化某一方面。希望更多新加坡人能理解，我们必须在速度、安全、成本和信任等各方面进行优化。
主席：谢谢国务部长。关于此事，沙拉尔·塔哈先生，您是否愿意撤回您的修正案？
沙拉尔·塔哈先生：我本想再问一个补充问题，但我会听从您的指示，主席。
主席：不，不，不。[笑声]
下午6时
沙拉尔·塔哈先生：好的，主席，我借此机会感谢张燕菁部长、高级国务部长陈杰厚、国务部长拉哈尤、劳明霞国务部长以及MDDI常任秘书和整个MDDI团队的回应和出色工作。我还要感谢我们的GPC和各位议员提交了25项削减建议，并进行了超过三小时的辩论，事无巨细；我相信我们中还有不少人想再问几个补充问题。
人工智能在本预算中占据重要位置。人工智能不是传统的政策领域，它的发展不是以十年计，而是以月甚至周计。前沿模型在短周期内容量翻倍，具备自主决策能力的代理系统正在出现，地缘政治争端影响——
主席：沙拉尔·塔哈先生。
沙拉尔·塔哈先生：——能源、数据流和供应链。然而，在这种模糊和高速变化中——
主席：沙拉尔·塔哈先生。
沙拉尔·塔哈先生：——MDDI必须同时兼顾多项任务，推动创新，维护信任，培养深厚且广泛的技能，保护弱势群体，确保国家竞争力。这绝非易事，我感谢MDDI团队所做的一切。基于此，主席，我请求撤回我的修正案。
[(程序文本) 修正案，经许可，撤回。 (程序文本)]
[(程序文本) 2,993,365,900元拨款用于Q项，列入主要预算。 (程序文本)]
[(程序文本) 119,025,200元拨款用于Q项，列入发展预算。 (程序文本)]
英文原文
SPRS Hansard 原始记录 · 抓取日期：2026-05-02
The Chairman : Head Q, Ministry of Digital Development and Information (MDDI). Mr Sharael Taha.
3.17 pm
Enabling AI as Strategic Advantage
Mr Sharael Taha (Pasir Ris-Changi) : Thank you, Mr Chairman. Mr Chairman, I move, "That the total sum to be allocated for Head Q of the Estimates be reduced by $100".
Mr Chairman, artificial intelligence (AI) features prominently in this Budget and this House has long recognised its importance. From a National AI Strategy in 2019 to a National AI Strategy 2.0 (NAIS 2.0) in 2023, we have moved from experimentation to scaling. The MDDI Government Parliamentary Committee (GPC) has consistently championed across six key themes. Our cuts today reflect coordinated scrutiny.
First, strengthening Singapore's AI value proposition. My cuts would seek clarity on our global competitive edge and the role of the National AI Council while Dr Choo Pei Ling and Ms Jessica Tan will press on delivering measurable outcomes, not just technical activity.
Second, building deep and broad digital capabilities. Mr Henry Kwek, Dr Choo Pei Ling, Ms Cassandra Lee and I will deliver cuts on how we are scaling AI skills across our workforce and enabling multinational enterprises (MNEs) and also small and medium enterprises (SMEs) to adopt AI meaningfully.
Third, ensuring ethical digital governance. Mr Christopher de Souza, Ms Jessica Tan, Ms Tin Pei Ling and I will be delivering cuts on the regulatory safeguards and accountability frameworks to ensure ethical and responsible growth, particularly as we move towards more autonomous, agentic AI systems.
Fourth, inclusive growth and uplifting vulnerable groups through technology. Ms Cassandra Lee and myself will deliver cuts on creating opportunities for fresh graduates, youth, seniors and lower-income groups so technology expands opportunity.
Fifth, investing in infrastructure and cybersecurity. Ms Jessica Tan and I will examine how we are strengthening our cybersecurity posture amidst increasingly sophisticated AI-enabled threats.
Sixth, building a high-trust digital society. Ms Tin Pei Ling and Ms Jessica Tan will raise cuts on trust, safety and protection from scams, deepfakes and online harms.
These cuts reflect our GPC that has been deliberate, aligned and persistent in championing the work together with the Ministry to secure Singapore's digital future.
Mr Chairman, allow me to begin on my first cut on strengthening Singapore's AI value proposition to the world and remaining relevant to the global market.
Mr Chairman, the global AI race has accelerated dramatically. At the recent Global AI Summit, it was clear we have entered an era of frontier and foundational models powered by massive compute and multimodal capabilities.
Generative AI (GenAI) is no longer experimental. It is embedded in enterprises, public services and national systems. The shift towards agentic AI and AI-native enterprises signals structural transformation.
This transformation is unfolding at three levels. First, population scale. In China, AI is integrated across platforms serving hundreds of millions. In the United States (US), AI co-pilots are embedded in productivity tools used today globally. In India, AI is woven into telecommunications and digital services at a national scale. AI is now part of daily workflows.
Second, compute scale. The race is not just about compute dominance. Next generation chips are being ordered at unprecedented volumes. Hyperscalers are investing billions in data centres. India has announced ambitions to attract up to US$200 billion in AI and data centre investments while the Middle East and the European Union (EU) are securing sovereign compute capacity. Chips, data centres and energy are now strategic infrastructure.
Third, industrial scale. AI is embedded in manufacturing, logistics, defence and energy systems. This is about industrial competitiveness and national capability. As capability accelerates, responsibility must keep pace. Safety alignment evaluation and red teaming are essential. Trust will determine who can scale.
Singapore cannot compete on population or compute scales, but we can compete on precision, trust, regulatory credibility and deep sectoral integration. This is not just about technology. This is about jobs and national competitiveness.
What is Singapore's unique value proposition in the global AI landscape? How do we compete against the population scale, computing scale and industrial scale of AI giants, like the US, China and India? How do we leverage on our high-trust governance, regulatory credibility and deep sectoral concentration?
Hence, I welcome the announcement of our National AI Council.
For it to succeed, clarity of mandate and the ability to execute are critical. What precisely is its role? Will it have execution authority, oversee cross-Ministry implementation, or remain advisory?
In a fast-moving AI race, structure must lead to decisive action. How will the council integrate economic strategy under the Ministry of Trade and Industry (MTI) with digital governance under MDDI to ensure coordinated delivery?
The National Artificial Intelligence Strategy 2.0 (NAIS 2.0) was launched in 2023. What progress has been made? Will the council build on and strengthen it, rather than duplicate or dilute existing efforts? How will interagency friction be resolved when priorities compete?
Finally, how will the council stay connected to industry realities? Will the industry leaders support it and be part of the council? Will SMEs have a meaningful voice at the table?
Thus, I am supportive of a concerted effort to focus resource on our national AI Missions and have champions of AI to accelerate AI deployment. However, I would like to seek clarification from the Ministry.
How did the Government determine the four key sectors for the national AI Missions? What criteria were used? What defines the success for these missions? It must go beyond pilots to measurable, real economic and societal outcomes. As we move at pace, when will we consider expanding into additional missions? I hope they are in months.
Finally, how do we define and measure success for our AI champions in terms of capability, adoption, global competitiveness and real impact on jobs and productivity?
On my last point. For us to be relevant in a global AI supply chain, we must prepare for compute infrastructure.
What is our plan to develop more compute capability? In light of the geopolitical tensions, we cannot rely solely on compute demand being supported predominantly from overseas. How much of sovereign AI compute capacity do we have? How do we be less dependent on others for compute capacity? Also, what is our clean or renewable energy strategy to support our AI-scaled data centres?
[(proc text) Question proposed. (proc text)]
The Chairman : Mr Sharael Taha. You may take your three cuts together.
AI as Strategic Advantage
Mr Sharael Taha : Thank you, Mr Chairman. If Singapore is to harness AI as a strategic advantage, three enablers must move together – deep and broad workforce capabilities, widespread enterprise adoption, including SMEs and strong ethical governance. The Government has introduced multiple grants, including Chief Technology Officer (CTO)-as-a-Service. What measurable progress have we made in accelerating AI transformation? Where are the gaps?
[Deputy Speaker (Mr Xie Yao Quan) in the Chair]
Unlocking AI's value requires more than just tools. It requires re-engineering business processes and redefining operating models. How are we helping enterprises integrate AI into core workflows and build AI-ready teams and deliver tangible productivity gains?
For SMEs, three constraints persist: cost of talent, lack of validated use cases and integration complexity. If AI adoption concentrates among only the large enterprises, productivity and wage gaps will widen for the SMEs. We must move from advisory support to shared capability infrastructure.
Hence, I propose three enhancements.
First, evolve our CTO-as-a-Service into AI capabilities as a service, where pooled AI engineers deployed across SME clusters can deploy solutions for hands-on implementation. Second, develop shared industry platforms to co-create plug-and-play AI models for common functionalities, such as quality control and logistics optimisation, to help reduce experimentation costs, especially for SMEs. Third, introduce outcome-linked co-funding tied to measurable productivity, export growth or energy efficiency outcomes. AI infrastructure should be treated like shared physical resources, enabling our SMEs to compete by agility and not by size.
Beyond enterprises, workforce development is also critical. What progress have we made in developing deep AI expertise through programmes, such as TeSA and in strengthening AI literacy across the broader workforce? How can AI widen opportunity for seniors, our fresh graduates and women returning to work? Can we create structured pathways for seniors to use AI tools meaningfully? And with entry level roles disrupted, how are we redesigning jobs and apprenticeship programmes so graduates gain both technical and business context skills?
Finally, as agentic AI systems become more autonomous, how do we ensure ethical and safe deployment?
There are clear areas where AI is ethically beneficial – assisting doctors in diagnostics, for example, or detecting financial fraud or optimising energy consumption or supporting seniors with daily living. But there are also boundaries that must not be crossed, such as autonomous lethal decision-making in defence, manipulative behavioural targeting, opaque credit scoring that entrenches biasness or AI agents making employment decisions without accountability.
The challenge is compounded when agentic systems operate with limited transparency, evolving goals or emergent behaviours that even developers may not fully predict. Who ultimately bears responsibility when harm occurs? The developer, the one who deploys it or the operator? What governance framework, audit requirements or red teaming standards and explainability thresholds will be mandated to ensure trust keeps pace with capability?
Digital for Good
Mr Chairman, as we advance in AI and digital transformation, how do we ensure technology is truly a force for good?
First, how has the Government strengthened digital delivery of essential services, particularly for persons with special needs such as the visual impaired? Are our systems inclusive by design? Second, as our society ages, how can digital and AI solutions better support seniors while easing the burden on sandwiched generation families? Third, how do we ensure children from lower-income families gain access not just to devices but to AI skills and also opportunities? Finally, how do we strengthen trust in our digital space and better protect Singaporeans from scams and online harms?
Technology must not widen divides. It must uplift, protect and empower every Singaporean.
Strengthening Our Cybersecurity Posture
Cyber threats are no longer isolated incidents. They are persistent, adaptive and increasingly AI enabled. As AI systems become more autonomous, we must confront a new class of risks – AI agents that can plan, probe and act independently.
How is MDDI addressing threats where malicious actors deploy AI to automate reconnaissance, craft sophisticated phishing campaigns or exploit vulnerabilities at scale?
3.30 pm
Singapore has experienced breaches before. The 2018 SingHealth cyberattack compromised the data of 1.5 million patients. More recently, it was reported that 255 firms linked to Singapore's critical infrastructures were allegedly targeted and ransomware incidents have disrupted healthcare clusters and third party vendors. Last year, a cyber incident affecting a critical information infrastructure (CII) operator reminded us that our CIIs, from energy to transport, remain a prime target.
Advanced persistent threats (APTs) are patient, well-resourced and may be state linked. They are not looking for disruption alone but strategic leverage. How then are we strengthening our national cybersecurity posture to defend again APTs? Are we investing sufficiently in threat intelligence fusion, real-time monitoring and cross sector incident response?
And on CII, regulations must keep pace with evolving threats. How will the Government enhance cybersecurity requirements for CII operators? Beyond compliance, what tools, shared platforms and AI-driven detection capabilities are being provided to help operators defend against sophisticated attacks?
Equally important, how are we strengthening the security posture, not just of CII owners, but also of their vendors and cybersecurity service providers, given supply chain vulnerabilities?
Finally, cybersecurity is ultimately about people. How are we expanding and deepening our cybersecurity workforce to defend against today's threats? Are we accelerating specialist training, midcareer conversion and advanced AI security integration skills? As AI becomes both a threat factor and a defence tool, how are we supporting organisations to adopt AI-driven cybersecurity solutions responsibly and effectively?
In a world of escalating digital conflict, resilience is not optional. Trust in our digital economy depends on our ability to defend it. We must ensure that as Singapore digitises at scale, our cybersecurity posture strengthens at equal speed.
Cyber Defence
Mr Gerald Giam Yean Song (Aljunied) : Sir, on 9 February 2026, the Government revealed that Singapore's major telecommunications operators were targeted last year in a sophisticated cyber espionage campaign by the group UNC3886. Such intrusions are a stark reminder that the digital battle space has expanded into a theatre of strategic sabotage, APT actors, prepositioned malicious code to sit dormant for years, designed to be activated during a crisis to trigger power failures or disrupt transport and payment systems.
For Singapore, this poses a direct threat to our national survival as a coordinated disruption to civilian telecommunications, payment systems and transport networks would directly cripple the Singapore Armed Force's ability to mobilise and deploy troops at speed. While the containment of UNC3886 demonstrates our technical proficiency, we must leverage this capacity to signal clear consequences. The Government must work with international partners to communicate strategic red lines, explicitly stating that the prepositioning of malicious code in our critical infrastructure is an unacceptable provocation. We must leverage our attribution capabilities to call out such actors directly, while carefully weighing the diplomatic sensitivities of naming state-linked groups. We should move toward a posture of active deterrence through precise signalling and the threat of calibrated counter measures. By doing so, while remaining consistent with international law, we can avoid unintended escalation. Ultimately, we must effectively change the cost benefit calculus of any potential aggressor.
Quantum-safe Cryptography Solutions - Why Buck Global Consensus
Mr Kenneth Tiong Boon Kiat (Aljunied) : Sir, Singapore, sits outside global consensus. In January 2024, the cybersecurity agencies of France, Germany, the Netherlands and Sweden jointly assessed Quantum Key Distribution (QKD). Their conclusion: QKD is not yet sufficiently mature and can only serve niche use cases. They said: "Migration to post quantum cryptography (PQC) has priority over the use of QKD".
PQC first is not just an American position. The PQC algorithms were built by European researchers. The back-up algorithm is entirely French. Eighty-two candidates from 25 countries went through eight years of open cryptanalysis. Germany published PQC migration guidance in 2020, four years before standards were finalised. Australia's deadline to cease classical public key cryptography is in 2030; Japan, 2035; 18 EU states signed a PQC commitment last November. QKD was not mentioned. PQC is software – it deploys on current infrastructure. Apple shipped PQC to 1.3 billion devices with an iOS update. Google enabled it for 3.4 billion Chrome users. Cloudflare has protected 20% of global web traffic since late 2023. No new fibre, no specialised hardware. A software update.
Without PQC, adversaries harvest today and decrypt tomorrow.
Singapore's position is the opposite. We are expanding the National Quantum Safe Network with dedicated QKD fibre, yet have no PQC migration deadline. Only Singapore and China are scaling QKD as national infrastructure rather than treating it as a niche research pilot. Our flagship quantum spin-off sells QKD back to the Government that funded it.
I ask why the balance between QKD and PQC appears opposite to every comparable nation. Many quantum researchers in Singapore are sceptical. They deserve accountability. Will the Minister disclose how the quantum safe budget breaks down between QKD and PQC, and when Singapore will set a PQC migration deadline?
The Chairman : Ms He Ting Ru, you may take your three cuts together.
Indecent AI Content
Ms He Ting Ru (Sengkang) : Sir, the upcoming enactment of Online Safety (Relief and Accountability) Bill increases support for victims of indecent online content. The Infocomm Media Development Authority (IMDA) has been engaging with X over Grok's generation of non consensual intimate images that were distributed en masse on the X platform. IMDA said that X has taken measures to address the issue, including stopping Grok from producing such content. Even as we ensure that the operating environment for tech platforms is not overly restrictive, could the Government explain further the outcomes of its engagement with X? Were any punitive actions taken over the matter? After introducing the spicy mode feature, Grok rose to the top 25 apps in the free Singapore Apple App Store in January this year.
Secondly, I quoted in my intervention during last year's Committee of Supply (COS) that there were reports of students generating deep fake news of their classmates and sharing them in WhatsApp groups. We must thus tackle the real problem. The existing and increasing demand for sexualised images, which is exacerbated by accessibility.
Given that most victims are women and children, the increased accessibility puts further pressure on these groups. We must do much more to educate our youths on the usage of AI, especially if increased exposure to it, and as early as Primary 4. Given concerns about how children handle AI, how do the Ministry of Education (MOE) sexual education approach and AI framework cover the issue explicitly? And how does MOE negotiate students' emotional engagement with AI chatbots?
The relationship between these images and the development of young Singaporeans is especially relevant as platforms work to become more addictive. More concerns beyond our existing legislation may become pertinent, such as content that does not involve specific victims, but nonetheless have societal concerns, such as AI-generated child pornography.
Social Media and Children
Sir, a child doom scrolling past bedtime is not making a choice. They are responding to a system designed to make stopping almost impossible. The current age assurance assessment, the Online Safety (Relief and Accountability) Bill and the code of practice for online safety represent a concerted effort to protect children online. Today, I want to ask whether it addresses a distinction not yet resolved. The difference between content harm and design harm.
Singapore already understands this. Regulation of the Sentosa and Marina Bay Sands casinos builds deliberate friction through entry levies, exclusion orders and visit limits. This recognises the need for behavioural design interruption, not just better information about the risks. On social media platforms, infinite scroll, autoplay videos and algorithmic feeds are attention-capture dark patterns designed to maximise engagement by exploiting reward-seeking and eroding self-regulation in children whose brains are still developing.
Last month, the European Commission made a preliminary finding that TikTok's addictive design itself is a legal violation. TikTok is designated under our own code of practice, and the commission found that its screen time tools and parental controls do not effectively address these risks. Silence from Singapore adds a reputational risk. An article in Nature Health last week stated that we must hold platforms accountable for their addictive design. These platforms exploit children's brains and erode children's capacity for self-regulation. The question is therefore whether we should allow platforms to deploy attention-capture dark patterns against children without legal consequences.
Could the Minister thus clarify three things? One, does the code of practice require designated services to submit a design risk assessment, covering recommendation systems, auto play and scroll architecture? And does IMDA have power to act on those assessments independently of content classification? If so, will we commit to a timeline for doing so?
Two, given the findings about TikTok's addictive design, has IMDA reviewed TikTok's compliance report with this in mind?
Three, would the Ministry consider my call last week to use a Select Committee to better examine global efforts to protect children from the harms of social media, especially in the light of momentum building to outrightly ban social media for children? Both children and their parents deserve a framework that holds platforms accountable, not just for what they show, but how they are built. Digital environments do not shape themselves. They are designed. And design, when left unchecked, becomes policy by default.
Lessons from the Albatross Files
Sir, history is not one dimensional. It constantly awaits further completion with access to more information. Knowledge about the past is important in shaping how we think and act in the present in tangible ways. Declassification is critical to this process. More credible and independently verifiable information is crucial when disinformation, misinformation, confusion and uncertainty arrive. Transparency is not just for transparency's sake.
Recent access to the Albatross Files underscores that separation with Malaysia was by mutual agreement. This makes it more possible to go beyond the historical narrative of trauma surrounding being kicked out in Singapore, as we look to advance ties with our closest neighbour.
Elsewhere, opening the Epstein files enabled some of the richest and most powerful people in the world being held to account for wrongdoing, and led to figures like Andrew Mountbatten Windsor and Peter Mandelson to be arrested. Access and accountability are especially important when the concerned persons remain alive, regardless of whether an issue is particularly heinous or more mundane.
Minister Josephine Teo stated that when deciding on access to public archives, state agencies take into consideration, and I quote: "supporting research into our collective past while safeguarding sensitive information and complying with relevant confidentiality and other obligations".
We should add timelines, holding state agencies and political authority to public account and avoiding confusion as well as misrepresentation. We pledge to aspire towards democracy. In a democracy, state action needs to be defensible to the public it serves, and from which it derives funding. Publicly indefensible positions and actions should not be undertaken. Therefore, these decisions must be ready to stand to public scrutiny at any time. Knowledge of this possibility encourages greater prudence and responsibility.
The Chairman : Mr Fadli Fawzi. Please take your two cuts together.
AI and Media Literacy
Mr Fadli Fawzi (Aljunied) : Sir, as Budget 2026 advances Singapore's AI ambitions, we must confront that increasingly Singaporeans are exposed to AI-generated misinformation and AI-powered scams at unprecedented scale and speed. A 5 February article in Lianhe Zaobao documented a surge of sensational videos claiming that Prime Minister Lawrence Wong is being forced out and that intense internal power struggles are unfolding.
These videos are entirely generated using AI within minutes, at a cost reportedly as low as US$1 or US$2 per 20-minute video. MDDI acknowledged that it has observed multiple online accounts publishing such fabricated claims about Singapore's domestic politics. An MDDI spokesman quoted by Zaobao stated that public education measures and resources have been rolled out and urged the public to rely on official sources and refrain from sharing unverified content.
I welcome this response, but I wonder if these measures are sufficient. Given the scale and sophistication of AI-generated misinformation, why was the Protection from Online Falsehoods and Manipulation Act (POFMA) not used against those behind these videos? Enforcement tools like POFMA alone also cannot inoculate society against misinformation. We need a population equipped to question, verify and critically assess what they see online.
What structured long-term programmes will the Ministry develop to strengthen media literacy and critical thinking, especially among vulnerable populations such as seniors? Will we expand community-based workshops, school curricula and public campaigns that teach citizens practical verification steps, such as checking original footage, examining sources, and consulting authoritative channels? Can we leverage AI itself to help filter and flag suspicious content at scale?
If AI lowers the cost of deception to $1 per video, then the cost of inaction may be far higher. How will our national AI strategy ensure that Singaporeans are empowered to discern fact from fiction in an increasingly polluted information ecosystem?
Declassification and National History
Sir, the recent declassification of the Albatross file has transformed out understanding of Separation. For decades, the official narrative surrounding Singapore's independence was that we were abruptly and unilaterally expelled from Malaysia by the Federal Government. That story has shaped how generations of Singaporeans understand our nation's founding.
Yet, the Albatross File departs from the prevailing narrative. The documents reveal that after the racial riots in July 1964, confidential talks had already commenced between the PAP and Malaysia's Alliance Party, regarding possible constitutional rearrangements within Malaysia. These discussions eventually led to separation.
3.45 pm
This revelation does not diminish our history, it shows that history is often more complex than we think and enriches our understanding of history. But why did this historically significant recourse take so long to come to light? And how many other important records remain inaccessible?
The recent declassification of the Albatross File illustrates why a Freedom of Information Act and automatic declassification is necessary. The Workers' Party (WP) has long called for a freedom of Information Act, most recently in our General Election 2025 manifesto. This call is grounded in a simple principle – we trust Singaporeans with the information necessary to hold the Government accountable.
Citizens should be empowered under a Freedom of Information Act to make requests for, and given access to, information from public agencies at the level of detail that is requested. Any data or records held by the Government that could inform public debate should also be automatically declassified and made available to the public after 25 years, subject of course to legitimate national security concerns.
Without a Freedom of Information Act and a framework for automatically declassifying record after 25 years, the Government would not be compelled to review and release information, and foundational truth risks remaining buried, indefinitely.
Facts and declassified documents can be cherry-picked to support a curated narrative. This is not what we want for Singapore. A Freedom of Information Act and automatic declassification would shift the burden of proof from citizens to the Government. The Government must justify why secrecy is needed, instead of citizens justifying curiosity. This empowers historians, journalists, civil society and ordinary Singaporeans to scrutinise decisions made in their name.
A mature nation does not fear its own archives. I believe that what the WP has proposed would strengthen our national identity rather than weaken it. We build a genuine national identity when it is grounded in facts, even when these facts are complex or uncomfortable. If we truly believe in accountable governance and informed citizenry, then it is time to enshrine the public's right to know, in law.
The Chairman : Mr Christopher de Souza, please take your two cuts together.
Immense Value in Historical Exhibitions
Mr Christopher de Souza (Holland-Bukit Timah) : A core responsibility of MDDI is shaping the historical narratives for Singaporeans and ensuring that our national historical touchstones are accurate, principled and grounded. In this regard, I wish to place on record my strong support for the recent exhibition on the Albatross File and the team of officers behind it. They should be commended.
The exhibition does excellent work in bringing into sharp focus the circumstances of independent Singapore's birth, the pressures of the time – 1963 to 1965 – and the difficult trade-offs faced by our leaders.
Understanding this history is not an academic exercise. The exhibition functions like a compass. It shows how decisions were made firmly on principle, but tempered by deep pragmatism. Faced with existential uncertainty, our Pioneers dug in, made hard choices and built a nation against the odds. These decisions made in the crucible of crises, shaped the DNA of independent Singapore.
There is scope for similar exhibitions. For example, on COVID-19. The pandemic was, after all, a recent and defining moment in Singapore's post-independent history. Those late-night decisions, vaccine procurement choices, protecting lives and livelihoods, and the multiple debates we had in this House, between 2020 and 2022, and resuscitating Singapore Airlines. It was a crisis of a generation and we prevailed together.
Decisions made in past crises shaped the DNA of a nation, and provide the ballast and compass for Singapore's future. The decision-making process should be displayed. Lessons learned from them in more exhibitions, such as that on the COVID-19 crisis.
AI - Discernment while Innovating
Sir, AI presents clear advantages. It can consume and summarise vast bodies of knowledge, but ultimately, AI is a tool. It cannot be allowed to be the master. It does not moralise.
Thus, as we embrace AI, we must do so with discernment. We should use it to the extent that it facilitates decision-making, but it cannot be allowed to usurp our decision-making.
Innovation cannot come at the expense of trespassing on some existing IP rights. And here, Sir, allow me to declare that I am a partner in a private law firm, practising IP law and advising on AI law.
Guardrails matter. In my view, there must be clear out-of-bound markers, no deepfakes, no deepfake pornography, no scams, no deception, no misrepresentation and no trespass on certain protected IP. If Singapore can be an engine for AI adoption while retaining its status as a trusted IP hub, we would have struck the right balance.
In short, to discerningly balance AI's benefits without atrophying the human mind, we must ensure that AI remains the tool, not the master.
Redesign Entry-level Jobs in Age of AI
Mr Abdul Muhaimin Abdul Malik (Sengkang) : Mr Chairman, the Government has committed a large amount of money to support industries in their AI transformation. In this context, I am concerned about how today's fresh graduates are already being impacted, and the long-term compounding implications this has on our national talent pipeline.
In a survey of 250 local employers, 80% acknowledged that AI has already reduced their entry-level hiring. As Members from both sides of the House have spoken about this problem, I will not belabour the point further.
Today, I would like to provide a proposal which I hope the Government will consider. From the Enterprise Innovation Scheme to the Productivity Solutions Grant, in exchange for the Government's support, stronger guardrails need to be established to prevent this support from accelerating the displacement of entry-level ones. Broadly, I suggest two conditions to be added for all AI-related roles and subsidy schemes.
First, at the minimum, companies should be required to submit a structured declaration detailing how their AI transformation efforts are expected to impact HR decisions, particularly for entry-level roles. How many such roles will be eliminated or redesigned, and what career development support will be made available to affected employees? This declaration will serve two purposes – to prompt companies to consider shaping their transformation efforts to protect their own talent pipeline and provide the Government with important insights on how entry-level roles may be impacted qualitatively and quantitatively across industries.
Second, companies should be required to commit to a sustained level of entry-level roles and ensure structured development opportunities for entry level highest. To keep compliance manageable for SMEs, this second condition could be made mandatory only for companies above a particular threshold.
To operationalise this, the National AI Council could coordinate with relevant agencies and consult the industry. For example, MDDI could take the lead in verifying companies' compliance with the conditions. In turn, only companies who receive this verification may submit Enterprise Innovation Scheme claims related to AI expenditures.
Sir, I believe my suggestions are practicable and necessary to ensure that publicly-funded support for AI transformation does not come at the cost of our national talent pipeline in the long term. Similarly, conditions are already in place for some subsidy schemes. The Productivity Solutions Grant already requires companies to submit a description of the overall impact of their proposed solution and specify the expected productivity gains before the grant application is reviewed and approved.
To conclude, the entry level jobs of today shape the industry leaders of tomorrow. Let us ensure that our AI transformation amplifies, rather than erodes the career opportunities that our young graduates have worked hard to earn.
De-risking AI and Automation
Mr Mark Lee (Nominated Member) : Chairman, Budget 2026 rightly places AI and automation at the centre of enterprise transformation. But if we want broad-based adoption, clarity and commercial realism matter as much as funding.
Many SMEs are not short of ambition, they are short of certainty. Uncertainty about what qualifies as AI or automation expenditure; how bundled digital costs are treated; how robotics hardware and software layers are classified; and what documentation withstands audit scrutiny?
When definitions are ambiguous, firms hesitate. In a tight cashflow environment, hesitation becomes inaction. Would MDDI work with MTI and other agencies to ensure that "AI and automation expenditure" is defined clearly in operational terms?
Without clarity, we risk two outcomes. First, AI-washing. Spending labelled as AI without measurable productivity impact. Second, under-adoption, with firms delaying genuine transformation due to compliance risk. Both weaken credibility.
Clarity alone, however, will not shift behaviour. For SMEs, the issue is risk asymmetry. Integration, deployment, robotics installation and workflow redesign costs are immediate. Productivity gains are gradual and uncertain.
If we want transformation beyond leading enterprises, the model must be simple – de-risk early, reward outcomes strongly. Would the Government consider strengthening upfront support to meaningfully reduce early-stage exposure, and then introduce performance-linked incentives where firms that demonstrate sustained productivity gains receive enhanced support, potentially up to 80% to 90% of qualifying transformation costs?
This would not subsidise spending. It would subsidise results – measurable improvements in output per worker, value-add per employee or cost efficiencies. Such a model aligns public spending with real productivity gains and gives SMEs the confidence to commit.
We must also be careful about how we frame AI nationally. Public discourse often centres on GenAI and digital tools. But in labour-intensive sectors – logistics, F&B, facilities management, manufacturing – robotics and advanced automation may deliver more immediate productivity gains. MDDI plays a critical role in shaping that narrative. Transformation is not just dashboards and chatbots. It is robotics, process redesign and job redesign.
Finally, coordination matters. AI-related schemes span multiple agencies. From an SME's perspective, the landscape can feel fragmented. Would MDDI consider strengthening a unified communication architecture, so enterprises see one coherent transformation pathway, rather than through multiple agencies?
In a structurally tight labour market, productivity is existential. Transformation must therefore be: clear in definition; coherent in communication; and commercially rational in incentive design. If we get this right – de-risk early and reward real outcomes decisively – we can achieve economy-wide transformation, not isolated pilot processes.
AI Opportunities for Growth
Dr Choo Pei Ling (Chua Chu Kang) : Mr Chairman, in my work with stroke survivors, we use brain imaging and machine learning to understand how the brain reorganises after injury. A single scan can produce thousands of images. Algorithms help us detect patterns we would otherwise miss. But no responsible scientist relies on a model blindly. We validate it rigorously, test for bias and examine when it breaks, because a wrong conclusion does not stay in a journal. It affects a person.
As Singapore accelerates our AI ambitions, we should bring that same discipline to national deployment. Budget 2026 sets a clear direction: a National AI Council chaired by the Prime Minister and national AI Missions to drive real outcomes across advanced manufacturing, connectivity, finance and healthcare. This is the right posture: AI – not as a buzzword, but as an economic strategy.
To make AI translate into growth that Singaporeans can feel, three disciplines matter.
First, value capture, not just adoption. Budget measures, such as expanding the Enterprise Innovation Scheme to support qualifying AI expenditures can spur uptake. But uptake is not impact. The jump from "trying tools" to "redesigning work" is where productivity is won. We should help firms, especially SMEs, cross that gap with mission-linked sector playbooks, reference workflows and practical benchmarks. AI must generate enterprise value, not just technological activity.
Second, trust architecture as a competitive advantage. In a fragmented world, Singapore's brand is that we are a place where serious systems run reliably. As AI systems move from assisting decisions to shaping outcomes, assurance cannot be informal. For high-impact deployments, institutionalising testing, explainability where needed and independent review where appropriate will strengthen confidence without stifling innovation. Trust is not a by-product of innovation. It is an asset we build deliberately.
4.00 pm
Third, bilingual talent at scale. We will need more than AI engineers. We will need people fluent in both domain and data, professionals who understand context, model limits and risk. The future workforce must be fluent in both code and context. If we build value, trust, and bilingual talent, Singapore will not merely adopt AI. We will shape how it is deployed and ensure that our growth is resilient and inclusive. I welcome the Minister's reflections on how MDDI will drive these disciplines through the National AI Council and the AI Missions.
The Chairman : Ms Cassandra Lee, please take your two cuts together.
AI-ready SMEs for Young Professionals
Ms Cassandra Lee (West Coast-Jurong West) : Mr Chairman, this AI transformation must be managed carefully because it brings anxiety to many.
From my conversations with youths, two concerns have been repeatedly raised. First, displacement by AI. Second, right-sized AI adoption.
On displacement, young professionals and youths preparing for their first jobs are concerned that their jobs will be displaced by AI. Youths I have spoken to have shared that they are worried that they cannot keep up with the pace at which AI is evolving. And as AI make some tasks redundant, their jobs may also be made redundant.
This applies to both youths looking to enter the workforce and those already in the workforce. Despite being digital natives, professionals and even technology professionals, they are nervous that it would be difficult to get ahead of the AI curve.
Many of those in the workforce are willing to upskill but they are time-poor. They need flexibility in training, employer support and buy-in and clear outcomes from each course that they take. Employers, in turn, ask for confidence and assurance that training translates into productivity.
So, the question is not what to train, but how to make training work in practice. What are the Ministry's plans to equip the workforce with the relevant confidence and skills needed to leverage AI in their respective domain expertise? What are the Ministry's plans to ensure that workforce training can be closely tied to improved productivity and business outcomes? What are the Ministry's plans to encourage greater employer buy-in and support to facilitate employee training? I support the Government's plans to explore how it can broaden the TeSA programme to help all young Singaporean workers continue to stay relevant.
On right-sizing AI adoption, AI adoption will not be one-size-fits-all. Different firms in different sectors face different constraints. This is especially true for our SMEs. As we all know, SMEs employ the majority of our workforce, with nearly half in small and micro enterprises. We cannot afford to leave them behind. But many SMEs face real constraints: cashflow, uncertainty of returns, manpower and implementation capacity.
So, support must be right-sized and practical. Not just funding, but end-to-end support to help enterprises adopt, integrate, scale AI in their core processes, redesign legacy systems and tailor solutions to different business needs. These uneven application of AI adoption heighten uncertainty amongst youths as to job security and progression.
In particular, I ask: how will the Ministry reduce uncertainty for SMEs adopting AI? For example, will the Ministry facilitate the provision of shared solutions or proven use cases jointly developed with trade associations and Institutes of Higher Learning?
The newly announced Champions of AI programme will go some way to support the integration of AI into business processes. How will the Champions of AI programme sit alongside schemes like the Enterprise Innovation Scheme and the Productivity Solutions Grant, given that they are administered by different statutory boards under different Ministries?
Renewed Possibilities for Libraries
I understand that MDDI is looking at supporting our libraries with renewed possibilities. I would like to request that the Ministry look at renewing our libraries with family at the top of its mind. Sir, I noticed I have run out of time.
The Chairman : Thank you. Mr Henry Kwek, you can take your two cuts together.
Spurring AI-centric IT Development
Mr Kwek Hian Chuan Henry (Kebun Baru) : Mr Chairman, AI-centric IT development is no longer theoretical. In Silicon Valley, leading AI firms and hyperscalers have moved beyond traditional software development. Frameworks now include intention-based engineering, vibe coding and agentic development. More top programmers are publicly saying they no longer code traditionally. AI-centric development is fundamentally different. It has integral to design, coding, testing and continuous improvement.
Singapore's IT services firms are the delivery layer between our national AI strategy and real-world outcomes. The bottleneck is not talent. Our tech workers are ready. It could be structural inertia within IT services firms and if they do not adopt these new paradigms quickly, our ambitions remain on paper.
Can MDDI work with the top AI firms and hyperscalers already based here to transfer such know-hows to our local firms? Should we leverage on these partnerships not just for enterprise development, but also to transform how our IT companies build software?
Can the Government Technology Agency of Singapore (GovTech) also move quickly to embrace these approaches, while adhering to cybersecurity and regulatory requirements, and progressively require IT service companies involved in Government work to do the same? This is not without precedent. We mandated Building Information Modelling adoption in construction procurement, and it transformed that industry.
Finally, how can we ensure our tech workforce and students keep pace? And beyond the tech sector, how can we encourage both our multinational companies (MNCs) and SMEs to embrace AI in their own operations?
Supporting Our National Media
Mr Chairman, our Public Service Media companies – Mediacorp, Zaobao, CNA, Business Times – are not just Singapore's truth infrastructure. They are also our society's trust infrastructure. In an age of AI-generated disinformation, like what some of our Members talked about earlier, they stand between our people and in a manipulated information environment.
Yet, our Public Service Media face significant headwinds, declining circulation in a fragmented media space, a rapidly evolving advertising model and growing disinformation from overseas. Unlike commercial outlets, our Public Service Media also carries nation-building obligations, serving all communities fairly, building social cohesiveness and upholding our national interest. Even reputable international outlets like the Washington Post have to resort to dramatic cuts.
Can MDDI outline its vision for keeping our national media compelling, relevant and thriving? How does it do so to help them stay financially viable, so we can forestall similar painful restructuring here?
We should not underestimate what we already have. Zaobao is already one of the most respected Chinese-language outlet globally. CNA commands credibility far beyond our shores. And with the strategic investment, The Business Times could become the Financial Times of Southeast Asia. Our Public Service Media a key source of our soft power.
Our media must also stay relevant for all Singaporeans, starting with our students. Countries like Australia and the UK have ensured that public service content remains prominent and easily discoverable on connected television platforms. Can we do the same, so that the quality of local content is not buried by algorithms favouring overseas programming?
Our Public Service Media companies are our national assets. I hope MDDI can share how it plans to secure our future. I notice I have 18 more seconds, so I want to add in a final point. I noted we are asking MDDI to do a lot of cybersecurity, national Public Service Media on IT services and AI development. I know that there is limited budget. I know it is not easy task. So, thank you in advance.
The Chairman : Ms Tin Pei Ling, please take your two cuts together.
Trust in a Digital World
Ms Tin Pei Ling (Marine Parade-Braddell Heights) : Chairman, we are an open society, physically and virtually. Information pours in from every direction and discerning truth has never been harder. This problem is now amplified by AI.
Most recently, many of us would have read about the fake reports online claiming that Senior Minister Lee publicly disagreed with Prime Minister Wong. A resident of mine whom I met during a block visit in January believed the story so wholeheartedly that I found it hard to dissuade him. Incidents like this do not merely misinform. They corrode mutual trust, weaken social cohesion and create fertile ground for scams. AI worsens the threat because it can generate convincing content at scale, iterate rapidly and be used to probe and undermine our critical information infrastructure.
Our Public Service Media plays a central role in preserving factual public discourse and remains the go-to source of truth for important issues by our citizens. Therefore, I have a few questions.
First, how will the Government support and strengthen our Public Service Media so it can more effectively in countering fake news and misinformation in an increasingly noisy information environment? This means funding, talent development, editorial independence and technical capability to verify rapidly and at scale.
Second, what concrete steps will be taken to ensure Public Service Media content remains high quality and highly accessible, across languages, platforms and demographics, so that credible information reaches every community before falsehoods do?
Likewise, how will we ensure genuine and verified Singapore narratives reach international audiences, both to protect our reputation from falsehoods and to project our voice on matters of global significance?
Third, will the Government equip our Public Service Media and our public agencies with advanced tools, including responsibly governed AI to detect, attribute and counter disinformation? Put simply, can we use AI to fight AI, with safeguards to avoid overreach, bias or erosion of privacy?
Finally, beyond Public Service Media, what broader investments in public digital literacy, rapid-response verification labs and partnerships with platforms and civil society will the Government make to build societal resilience to AI‑driven misinformation?
In an age when technology amplifies both benefit and risk, we must ensure our public information architecture is robust, trusted and adaptive.
AI Governance and Agency
As Singapore embraces AI as a strategic necessity for our development, we must protect the long‑term interests of our people even as we harness its power. Much of the public conversation today veers toward doom‑laden predictions of AI "taking over" jobs and society. That narrative overlooks something fundamental. We can and must retain human agency. We can choose how AI is designed, governed and deployed.
That choice demands strong national leadership to chart practical governance pathways and sustained international cooperation to agree what ought not to be delegated to machines. Singapore has led with the Model AI Governance Framework of January 2019, a world first, and followed up with work on Generative and Agentic AI. These are important foundations.
But leadership must translate into concrete action. Interoperable standards and certification, robust procurement and audit requirements, independent oversight and investments in public literacy and workforce reskilling so that citizens can exercise meaningful agency. Internationally, we could forge norms that prevent regulatory arbitrage and ensure cross‑border accountability.
Hence, building on existing frameworks, what more will the Government do to strengthen Singapore's role in global AI governance, standards, certification, international coordination and capacity building, to preserve human agency while responsibly constraining AI agency?
The Chairman : Ms Jessica Tan, please take your three cuts together.
Digital Safety and Societal Resilience
Ms Jessica Tan Soon Neo (East Coast) : Mr Chairman, Budget 2026 underscores an important priority, keeping Singaporeans safe from scams and online harms. Digital safety today is no longer just about avoiding suspicious links. AI has changed the way scams and online threats work. The tools used to deceive people are more sophisticated, more personal and harder to detect, even for those who are usually confident online.
We now see AI‑generated deepfakes that could sound exactly like a family member or friend with uncanny accuracy. Hyper‑personalised scams tailor messages to a person's habits, vulnerabilities and online behaviour. Misinformation spread faster at a scale that outpace fact‑checking. And these risks fall disproportionately on seniors, youth, lower‑income families, and those who may not have the digital confidence to tell what is real and what is AI-generated.
To keep Singaporeans safe, we now must move from "digital safety" to AI risk resilience, equipping people with practical skills, trusted tools and strong community support.
We can strengthen this in four ways. I would just like to suggest these.
One, introduce a national AI safety curriculum across digital literacy programmes for different age groups and life stages.
Two, the Online Safety Commission (OSC) can incorporate AI-specific risks into its categories of online harms by recognising AI-generated impersonations such as deepfakes and mass production of inauthentic materials to enable victims to seek timely remedies.
4.15 pm
Three, establish a public-private AI anti-scam taskforce to stay ahead of evolving threats and coordinate responses.
Four, leveraging the base of Digital Ambassadors, develop a network of community-based digital safety ambassadors focused on AI risks in senior activity centres, schools and social service settings.
Mr Chairman, digital safety tools are about protecting trust, protecting families and ensuring every Singaporean, regardless of age or background, feels confident and safe in an AI-driven world.
Responsible AI
Mr Chairman, this year's Budget rightly emphasises the importance of responsible AI. Singapore has already built strong foundations – from the Model AI Governance Framework to AI Verify – and sector‑specific guidelines in finance and healthcare. These are real strengths, and they show our commitment to safe and trustworthy innovation.
But as AI becomes part of everyday decisions that affect people's lives, Singaporeans now need more clarity, more consistency and more accountability.
Today, AI is already used in hiring, credit assessment, insurance underwriting and even public sector processes. But the level of transparency varies widely. Many Singaporeans may not even know when AI is involved. Without regular checks, these systems can unintentionally reinforce or amplify bias. Trust does not happen automatically. We must build it deliberately.
At the same time, we know that modern AI, especially frontier models, is complex and often proprietary. While transparency and independent evaluation sound simple, the reality is more challenging. Singapore's existing frameworks recognise this, but the pace of deployment means we need to strengthen our approach in a practical and proportionate way.
Not all AI systems can carry the same level of risk. A chatbot answering Frequently Asked Questions is not the same as an algorithm screening job applicants, assessing creditworthiness or supporting medical decisions. And frontier AI models, the most powerful and unpredictable, pose a different category of risk altogether.
That is why I believe Singapore should move toward targeted requirements for high‑risk or high‑impact AI systems rather than broad, one‑size‑fits‑all. And transparency does not mean revealing source code. It simply means explaining what the system does, what risks it carries and what safeguards are in place. Independent audits should only be required where the potential harm is significant.
These are not radical ideas. They are becoming global norms. The European Union already mandates audits for high‑risk systems. Canada is moving in the same direction. US regulators require audits in finance and healthcare. The United Kingdom (UK) is strengthening evaluation requirements for frontier models. Singapore should stay ahead, but in a way that fits our context and supports innovation.
A risk‑based approach allows us to protect Singaporeans while keeping compliance practical. This ensures we do not overburden SMEs or slow innovation while still giving Singaporeans confidence that AI is being used responsibly.
My recommendation is for the Government to co-develop practical guidance, sandboxes and sector‑specific standards with industry, building on strong foundations we already have. This turns responsible AI into a shared national capability and not just a regulatory obligation.
Enterprise Readiness for AI Adoption
Mr Chairman, Budget 2026 gives Singapore's AI push real momentum. But for many SMEs, key questions remain. Will AI make daily work easier, more productive and more meaningful for our people?
SMEs still face real hurdles. Compute is costly, data is fragmented, governance feels complex and workers worry about job impact. If we do not address these realities, AI will benefit only a few.
The refreshed National AI Strategy sets the direction on trusted AI, but SMEs need practical tools that they can use tomorrow – sector‑specific AI trust roadmaps that spell out common risks and good practices, pre-approved governance templates for data handling, model testing and human-in-the-loop check, simple "green‑lane" guidance so low‑risk use cases can move quickly while higher‑risk ones get the safeguards they need. This is how trusted AI becomes a catalyst, not a compliance burden.
The Champions of AI programme is promising, but AI adoption is about more than tools. It is about preparing data, redesigning workflows and helping workers feel confident. Many SMEs lack this expertise.
Can MDDI share how SMEs can tap these champions for governance support, workflow redesign and to fully leverage national compute and enterprise schemes?
Our workers are central. The workforce transformation roadmaps must go beyond broad skills. Workers need role‑specific skill maps to show how tasks will change with AI, clear pathways to move from today's roles to tomorrow's AI‑enabled, and hands‑on training tied to tools that SMEs are actually adopting. When workers see how AI makes work easier and raises productivity, adoption becomes natural.
I welcome the Budget's investment in local AI developers and testbeds and the role of Government procurement in helping them scale. And as AI becomes more embedded in operations, the new Cyber Resilience Centre and enhanced SME support will give businesses the confidence to adopt AI safely.
Mr Chairman, when we combine practical support, clear AI trust guidance, empowered workers, strong cybersecurity and a vibrant local ecosystem, AI becomes a real productivity tool and boost for our enterprises – and a real competitive advantage for Singapore.
The Chairman : Minister Josephine Teo.
The Minister for Digital Development and Information (Mrs Josephine Teo) : Mr Chairman, I thank Members for their cuts. Let me start my response in Mandarin, please.
( In Mandarin ) : [ Please refer to Vernacular Speech .] Mr Chairman, in the blink of an eye, tomorrow will be Chap Goh Meh . Before Chinese New Year, I asked my mother if she wanted me to accompany her to buy new clothes, but the 83-year-old said, "No need! I have already found the clothes I like online and placed an order."
I was afraid that she might be scammed, so I asked her how she knew that the seller was reliable. She confidently replied, "I will only pay after I receive the goods and am satisfied with them."
On the day of our reunion dinner, she excitedly showed me her new clothes, and only then did I feel reassured.
Mr Chairman, digital technology has brought a lot of conveniences to our lives and created new opportunities for our businesses. However, it has also exposed us to unprecedented risks and dangers. Similarly, AI has both benefits and drawbacks. Several Members have also mentioned this.
Some Singaporeans worry that they cannot keep up with the pace of this AI era. I have also felt the same way before. However, as the Prime Minister said, we cannot stand still out of fear of AI.
As the saying goes, "Like a boat going against the current, you must move forward; otherwise, you will fall behind."
Other countries have developed their AI initiatives. If we do not act fast enough, plan broadly enough, or establish our foundations deeply enough, we will inevitably fall behind. The key is that our goals must be clear, and our measures effective.
In this AI era, how can we ensure that Singaporeans are not left behind and help SMEs maintain their competitive advantage? This is a core issue that we are closely monitoring.
Just like my mother – she is not a digital expert, but with appropriate help, she too can shop online safely.
We do not need to force ourselves to become AI masters, because not everyone can master AI to the same degree; the ways in which they benefit from it will also differ. More importantly, Singapore must remain confident so that we can move steadily ahead in this AI era.
In this year's COS debate, MDDI will propose various initiatives in this direction and ensure that Singaporeans can not only keep up but also benefit.
( In English ): Mr Chairman, AI has taken centre stage at this year's Budget and COS debates. Members have shared optimism about opportunities and anxiety over impacts on our jobs, creativity and autonomy.
Mr Sharael Taha asked a strategic question about Singapore's unique positioning in AI. We are fortunate that international counterparts recognise our ability to respond holistically across industries, enterprises and the workforce through a range of enablers – from R&D and infrastructure to safety and governance.
On the global stage, Singapore is frequently at the table. Our progressive, thoughtful approach to AI makes us a credible partner and useful reference point. This has made it possible to aim higher.
Prime Minister Wong, Deputy Prime Minister Gan and MTI colleagues outlined plans to grow AI champions and pursue national AI Missions. Later, Ministry of Manpower (MOM) and MOE colleagues will discuss how we empower the present and future workforce to make the most of AI. I will focus on what this means for the broader base of our businesses.
In gist, we want to take full advantage of AI's ability to be democratised, or to put it more simply, for its benefits to spread widely because solutions once too expensive or complex are more accessible.
But if AI follows the same path as previous technology waves, only a small group of companies at the frontier will get ahead and pull away from the pack. The long tail of smaller and often less-resourced businesses take much longer. Yet collectively, they employ most of our workforce. When they fall behind, more than GDP is at risk. At stake are our entrepreneurs' hopes and dreams, workers' livelihoods and their communities' progress.
This is why MDDI is creating the National AI Impact Programme – to turn AI's possibilities into reality for the many, not the few.
Today, 15% of SMEs and about seven in 10 workers use AI in some way. We want to encourage those who have not started to take the first step and help those already using AI move beyond basic applications.
Over the next three years, the National AI Impact Programme aims to support 10,000 local enterprises to integrate AI into their business processes. This will create a sizable pool of early adopters. They can be multipliers in the community, sharing their experiences and knowledge through the intermediaries that Ms Denise Phua asked the Prime Minister about.
Small businesses stand to gain the most. Take Durian Memories for example, a single store seller in Ang Mo Kio. They did not have the luxury of dedicating a team member to handle customer enquiries. Unsurprisingly, they lost sales when hungry durian lovers were not attended to.
But Durian Memories tackled this challenge by implementing an AI-enabled customer relationship management system with a chatbot that automatically answers customer queries. As a result, peak sales went up by 30%.
There are now many AI tools that improve business operations in simple, effective ways. They make up 30% of the digital solutions on IMDA's SMEs Go Digital platform today. We will expand the range of AI-enabled solutions with grant support to meet different business needs. More SMEs can then access these pre-approved, cost-effective and market-proven tools to integrate AI readily and affordably.
Like Mr Pritam Singh and Mr Muhaimin, we want these solutions to be transformative yet human-centred. At the same time, Mr Mark Lee worries about AI washing. We will put safeguards in place for grants and incentives whilst trying at the same time to not make the rules too onerous.
Some enterprises are ready to do more with AI. Take Mocha Chai Laboratories for example. They are a talented team of multimedia creators who improve film visuals and sound. Unknown to most of us, sound effects are still added manually to films, often taking four to eight weeks. After joining IMDA's Digital Leaders Programme and building up their tech capabilities, Mocha Chai created a new GenAI tool that analyses video footage and automatically generates matching sound effects, reducing weeks of work to just a day.
This innovation allowed the company to not only save costs but create a potential new income stream. It has opened up opportunities for both the business and their employees.
We want more success stories like Mocha Chai. But as pointed out by Ms Jessica Tan, Mr Mark Lee and Mr Sharael Taha, more sophisticated uses of AI require multiple factors to succeed. Often, the technology is ready, but people are not. This is why we are enhancing the Digital Leaders Programme and launching a new Digital Leaders Accelerator Bootcamp to build skills and confidence in change management and not just tech capabilities.
We also thank Mr Andre Low, Mr Dennis Tan and Mr Fadli Fawzi as well as Mr Sharael Taha for recognising the need to plan ahead, as the Government has done, to manage the energy impact of widespread AI use.
We do this in several ways. We are judicious in how we expand digital infrastructure. When allocating new data centres, we assess how well they use low-carbon energy sources. We are introducing new sustainability requirements to improve the energy efficiency of older data centres. And through the National AI Research and Development plan, we will support public research in resource-efficient AI to better understand our options.
4.30 pm
As more businesses adopt AI, there is also opportunity to uplift the workforce and help them stay relevant, whether at the entry level or at later stages of their careers. Beyond the Prime Minister's commitments and MOM’s plans, I want to assure Members like Mr Abdul Muhaimin, Ms Cassandra Lee and Dr Choo Pei Ling that MDDI is focused squarely on this.
We know that PMEs and knowledge workers feel the pressure more acutely. But many have found ways to be more effective with AI’s help. Take Geraldine Lau, an audit professional who has been with KPMG for 27 years. For each audit, Geraldine pores through reams of documents to assess risk. With employer-provided training, she created an AI agent that automatically consolidates key information from company announcements for audit reviews.
The AI agent organises information more quickly than Geraldine can, but her domain knowledge is key to ensuring it looks in the right places. With hours of manual work saved, she can now focus on deeper risk assessments and applying her human abilities – wisdom, calibration and professional judgement – to more complex work.
Geraldine and many PMEs are showing that AI know-how, domain expertise and human touch are a powerful combination. Not all of us can be AI engineers. But we can be “bilingual” in AI in our own areas of expertise and to solve problems in our domains.
For a start, the Government will support 100,000 workers to become AI bilingual. They will be pathfinders for meaningful AI upskilling, for others to emulate. Our initial focus will be on professions that are highly exposed to AI and serve multiple industries. IMDA will work with relevant agencies and professional bodies to expand its TeSA programme, to develop AI bilingual workers in key domains. We will start with the accountancy and legal professions, and extend our reach to other fields such as HR.
As Mr Henry Kwek noted, AI is also transforming the tech sector – many people can now write code and build prototypes with the help of AI. We will therefore enhance the TeSA offerings to help tech workers move up the value chain, from writing code, to orchestrating end-to-end systems powered by AI agents.
With AI evolving quickly, our governance must also keep pace. We agree with Ms Jessica Tan and Dr Choo Pei Ling on risk-based, practical AI governance. Like Mr Christopher de Souza, we believe AI should not replace the discerning human mind.
Our new Model Governance Framework for Agentic AI will help organisations manage systems that can act with greater independence, while ensuring human oversight. We are the first government worldwide to introduce such guidelines. For high-risk, high-impact systems like frontier models, we will progressively strengthen safeguards.
However, what we do locally is not enough, a point noted by Ms Tin Pei Ling. The most advanced AI models are developed in only a handful of countries, but their cooperation on AI safety is not deep.
In recent years, Singapore has hosted major AI conferences to promote international cooperation. Last year, we organised the Singapore Conference on AI: International Scientific Exchange on AI Safety. The exchange brought together world-class thinkers across research, government and civil society, resulting in the Singapore Consensus on global AI safety research priorities.
Recently, at the India AI Impact Summit, I shared that Singapore will host the second edition of the International Scientific Exchange to update the Singapore Consensus. Despite the challenges, we will continue contributing meaningfully to the international discourse on AI safety.
Next, on cybersecurity. Members are understandably concerned about whether our critical infrastructure is sufficiently protected against malicious threat actors, especially state-sponsored ones. I would like to reassure Mr Sharael Taha and Mr Gerald Giam that the Cyber Security Agency (CSA) works closely with domestic and international partners to detect and contain cyber threats.
On the diplomatic front, Singapore recently concluded our chairmanship of the Second UN Open-Ended Working Group on security of and in the use of info-comm technologies.
Realistically, state-sponsored threat actors are par for the course. It is nonetheless important to forge international consensus on what constitutes responsible state behaviour in cyberspace. We must, however, not expect these efforts to be a substitute for stronger cyber defense capabilities. In this regard, CSA will focus on three key areas.
First, we will review our cybersecurity standards and requirements for CII owners. Second, we will provide CII owners with advanced tools, so that they are equipped to deal with advanced threats and. Third, we will work with partners to build up capabilities in our cybersecurity workforce. Senior Minister of State Tan Kiat How will say more about these efforts.
Another risk we face is the spread of disinformation and misinformation, fueled by technologies like AI. As a diverse society, we are particularly vulnerable to online falsehoods that erode trust in our society and institutions. Fortunately, we have been strengthening our libraries and archives. They help to nurture a discerning population by cultivating reading habits and information literacy. Minister of State Rahayu will share more later.
Our Public Service Media entities too, are important in maintaining trust in our infospace. I thank Mr Henry Kwek and Ms Tin Pei Ling for recognising this. Our Public Service Media entities reach over 90% of Singaporeans. They remain highly trusted by the public, more so than reputable international and online media outlets.
Consequently, our Public Service Media entities have become indispensable in countering misinformation. MDDI will therefore continue working closely with our Public Service Media entities to maintain their reach and strengthen their fact-checking capabilities. For example, CNA will set up a digital verification team. Government agencies have also collaborated with The Straits Times on the AskST series to address misinformation.
Mr Henry Kwek asked about efforts to help Public Service Media remain relevant, discoverable and financially viable as audience attention and advertising shift towards digital platforms.
Besides delivering timely and credible news, our Public Service Media entities produce content that strengthens our sense of identity as one people. They also play a role in cultivating news literacy among our young, through regular student publications and school competitions.
Given the critical role of our Public Service Media, MDDI will support efforts to keep public service media content visible and easily discoverable. We are studying approaches in other countries and will consult the industry to ensure that initiatives are implemented reasonably and effectively. The Government will continue investing in our Public Service Media entities, helping them develop new capabilities as the media landscape evolves.
Sir, to conclude, the investments we make today will determine whether we lead or lag tomorrow. By accelerating AI adoption, strengthening technology governance, and building discernment among our people, we are positioning Singaporeans to seize the opportunities and make progress together.
The Chairman : Senior Minister of State Tan Kiat How.
The Senior Minister of State for Digital Development and Information (Mr Tan Kiat How) : Sir, we have made major moves in the last decade to shore up our cybersecurity such as setting up CSA and introducing the Cybersecurity Act to protect our critical information infrastructure.
But there is no room for complacency. I agree with Mr Vikram Nair’s cut to the Ministry of Home Affairs that threat actors, especially APTs, will only get more sophisticated. Mr Sharael Taha asked about the Government’s plan to protect our CII.
Cybersecurity is a collective effort. CII owners must take responsibility of the systems they own and operate. The Government will also do our part.
At this COS, I will speak about MDDI’s plans to first, update the cybersecurity standards and obligations; second, level up our CII owners; and third, strengthen capabilities in our cybersecurity workforce.
Today, our CII owners are held to higher standards and stringent obligations are imposed on their critical systems or CII systems. This was a calibrated approach to balance national security needs and business costs. We have observed that threat actors are also targeting non-CII systems because they may be less secured and can be entry points into CII systems.
CSA is therefore reviewing the scope of the current cybersecurity standards and obligations, and may include non-CII systems, such as networks that are interconnected with the CII systems. We are mindful not to impose unnecessary costs on CII owners, and will continue to take a risk-based, calibrated and pragmatic approach.
Sector Leads may introduce additional sector-specific obligations that are adapted for their sector. For example, IMDA will be enhancing its cybersecurity regulations for the telecommunications operators, given the recent waves of attacks. IMDA intends to provide guidance for areas such as managing virtualisation of infrastructure and credential management.
We expect CII owners to comply with these requirements. CII owners currently engage third parties to conduct audits and regular penetration testings to verify their robustness of their defences. These reports are then submitted to CSA for review.
In addition to relying on such third party reports, CSA wants to ensure that the security controls implemented by CII owners are not only tested and validated during audits but continuously strengthened. One way to do so will be to partner CII owners to do on-site reviews. CSA is currently discussing with the Sector Leads on the implementation plan. We will reach out to the identified CII owners when ready.
Sir, regulations and compliance can only go so far. We need our sectors and our CII owners to do their part to defend their systems, consistently every day.
Over the last year, I have visited the CII sectors, taking time to speak with the sector leads and the key CII owners. We have had closed door, candid discussions. Our Sector Leads and CII owners understand that the threat landscape has evolved and appreciate what is at stake. However, they shared with me that most CII owners are private companies whose business is in the delivery of essential services. They are not specialists in cybersecurity. Yet, they are up against the best-in-class, state-backed cyber threat actors. One of the Chief Information Security Officers told me that it is like he is bringing a knife to a gun fight. I empathise with his point of view.
As I said, cybersecurity is a collective effort. We are on the same team. Therefore, the Government will lean in to help CII owners to strengthen their defences and better respond to incidents.
Typically, national security is the exclusive domain of governments, such as developing cutting-edge technological systems and training skilled operators to deal with various threat scenarios. We have decided to avail some of the Government’s expertise to the private sector, to level the playing field between the defenders and the attackers. We will help our CII owners “level up” and hold their own in a fight against APTs.
First is intel. We will selectively share classified threat intelligence with our CII owners so that they are better able to spot and respond swiftly to threats that are attacking their systems.
Second is tools. We will equip CII owners with proprietary threat detection systems to strengthen their abilities to detect malicious activities in their networks, especially those of state-sponsored APTs. These proprietary tools complement commercial threat detection systems used by our CII owners today. We have started deploying these tools in selected CII owners and will progressively deploy them across the rest. CII owners may need to incur cost to integrate these tools into their systems. We will consider funding support, if needed.
4.45 pm
Even with these measures in place, we must be prepared that some threats will go undetected. This is why defenders must remain vigilant and constantly enhance their capabilities.
This brings me to my next point on innovation. Threat actors are also not standing still. As pointed out by Mr Sharael Taha, autonomous AI agents are emerging threats. We must similarly harness technology to defend our critical systems. CSA will partner with CII owners to test the use of technologies, such as AI, to help enhance their efficiency and effectiveness of their cybersecurity operations. We will share more details in due course.
The defenders will also need to be competent in using these tools effectively. Therefore, CSA will work with training providers to design and curate courses that equip cybersecurity professionals with specialised knowledge and skills on how to deal with APT threats.
The responsibility of securing our CII systems cannot just rest on the shoulders of our frontline cyber defenders. This is not just a technical matter. The Board and management of CII owners must also do their part. It is a leadership responsibility. We will equip them with the relevant knowledge.
Since 2021, CSA has partnered the Singapore Management University to conduct the Cybersecurity Strategic Leadership Programme for C-suite leaders. The programme has trained 74 senior leaders thus far, such as Ms Dewi Anggraini from SMRT, Mr Andre Shori from Schneider Electric and Mr Kang Seng Wei from DBS. In view of the participants' positive feedback, CSA will conduct more runs of the Leadership Programme over the next few years. We intend to welcome the next batch of cybersecurity leaders by the second half of this year.
Let me now turn to how we are protecting our citizens. Just last year, Members may have seen articles stating that attackers gained unauthorised access to thousands of Internet of Things (IoT) devices, including routers, around the world. Singapore has not been spared. Last year, attackers infected over 2,700 devices, such as baby monitors and routers. When such personal devices are hacked, citizens' privacy can be compromised and their daily activities being disrupted. These devices can also be used unknowingly to launch attacks against others.
The Government will do more to protect our citizens against these malicious actors. First, we will do more to ensure that the digital products that are sold in Singapore have baseline security safeguards in place. This will make these products harder to be compromised.
Today, we require home routers to meet minimum cybersecurity requirements. This is because they are the gateways to networks and transmit sensitive information. They are currently required to meet Cyber Labelling Scheme, or CLS Level 1. CLS is like the energy efficiency tick label you see on household appliances, but instead of showing energy use, it tells you how cybersecure the device is.
CLS ranges from Level 1 to Level 4, with Level 1 being the most basic standard. We have seen threat actors using more advanced techniques to exploit home routers. CSA and IMDA therefore intend to raise the minimum cybersecurity requirements for all routers sold in Singapore to the equivalent of CLS Level 2.
Besides routers, IP cameras are another common target for cyber threat actors. Threat actors exploit these cameras to spy on individuals. Exploited images are even uploaded onto pornographic websites or used to blackmail individuals. To better protect citizens, CSA will explore requiring IP cameras to meet CLS Level 2, similar to home routers.
CSA will continue to monitor and review if more digital devices should be required to meet minimum cybersecurity standards.
Second, for organisations which handle sensitive data, including personally identifiable information, we are considering to introduce more stringent cybersecurity and data protection obligations. The Government will take the lead in this. GovTech will require Government vendors that manage critical systems and sensitive government data to meet Cyber Trust Mark requirements.
CSA will also require the following three groups of entities who are operating, assessing or handling sensitive systems and data to meet Cyber Trust Mark Requirements. These are the CII owners, auditors conducting cybersecurity audits on CII systems and CSA's licensed Cybersecurity Service Providers providing penetration testing and managed security operations centre services.
Consultations with relevant stakeholders are ongoing and these measures will be implemented progressively over the next two years.
We are also looking ahead to prepare for tomorrow's threats. Mr Kenneth Tiong sought to clarify Singapore's approach to quantum-safe migration. We have been monitoring this technological trend closely. We also take the position that PQC will be the mainstream solution for quantum-safe migration. It is widely tested and internationally accepted. Singapore will take reference from the National Institute of Standards and Technology (NIST) standard as the baseline. As Mr Tiong pointed out, this is the position taken by many other countries.
QKD is a complementary technology. It is more for niche application, like securing high assurance communications. Singapore takes a risk-oriented approach when it comes to quantum-safe migration. The Government is reviewing the practical steps we can take for quantum-safe migration, including adoption of PQC and the appropriate role of QKD, if needed.
We have started investing in capabilities to support businesses in quantum-safe migration. In October 2025, CSA released a Quantum-Safe Handbook and Quantum Readiness Index to raise awareness of the associated risks. We are working with industry experts to better support organisations in their efforts, including through training. We are also deploying two quantum-safe networks nationwide through the National Quantum Safe Network Plus (NQSN+) initiative. This provides additional options for businesses to integrate quantum-safe solutions, such as PQC and QKD, into their networks and systems.
By supporting the provision of NQSN+ infrastructure and services, we aim to reduce the technical and financial barriers for organisations looking to implement quantum-safe solutions. Quantum-related technology is an evolving field. We are closely monitoring developments and will release guidance on this in due course.
We are prepared to adopt different technological solutions if they prove to be effective and able to meet our needs.
Sir, our digital infrastructure underpins our economy and daily life of citizens. MDDI is committed to improve the resilience and security of our digital infrastructure.
The Chairman : Minister of State Jasmin Lau.
The Minister of State for Digital Development and Information (Ms Jasmin Lau) : Mr Chairman, in today's world, many businesses provide seamless, effective and reliable digital services. Singaporeans expect the same from the Government.
While we have made good progress, we can do better. We must acknowledge that citizens still encounter services that are slower than they should be, forms that ask for information we already have or systems that do not speak easily to one another. My speech is about what must change.
Let me borrow an analogy from a familiar childhood toy – LEGO sets. There are a few things that we can learn from LEGO sets.
First, they are based on a strong understanding of their customers and how preferences change over time. Children grow. Attention spans evolve. LEGO designs have adapted accordingly.
Second, modularity. Every brick is designed to connect seamlessly with one another. Builders do not need to reinvent the basic structure each time. They reuse, recombine and build upwards.
Third, customers can choose how simple or complex a model to build. For some, a simple LEGO DUPLO or LEGO City set is sufficient. And for others who want complexity, LEGO Technic or LEGO Education SPIKE sets offer advanced mechanical components. These are optional but easily added on.
The analogy may sound simple, but the engineering discipline behind it is not. What do these lessons look like for us, in the Public Service?
We must understand our citizens continuously, not episodically. Singaporeans' expectations evolve. Their life journeys change. A service that felt intuitive ten years ago may feel slow or fragmented today. We must design modular systems and digital components that work across agencies. When systems connect seamlessly, citizens experience Government as a whole. And we must be able to provide services that meet complex needs. Our advanced shared tools, such as AI platforms and coding assistants, support us in building specialised components at scale. Not every problem has complex needs. But when complexity is needed, the capability must already be there. Secure, integrated and ready.
To do all of this well, some of our habits must change. Sometimes, digital transformation becomes a collection of projects. New apps and new pilots. Transformation is not about the number of digital projects launched. It is whether citizens find it simpler, faster and clearer to deal with the Government.
This requires discipline in problem definition. It also requires discipline in experimentation. In the Public Service, caution is natural. But doing nothing, while expectations move ahead of us, is a risk. We must learn to manage that risk, not eliminate experimentation.
Through initiatives, like Open Government Products' (OGP's) Hack for Public Good, our officers work closely with users to understand and address real pain points. A team observed that medical social workers spent long hours writing case notes after emotionally difficult conversations. Their first instinct was to build an automated transcription and summary tool. But the tool did not fully address the users' needs. Our social workers wanted greater control over how case notes were structured, so that important information can be retrieved easily. When the generated notes were not organised clearly, they rewrote them.
The team refined the tool into Scribe, an AI-powered tool that transcribes conversations and generates summaries according to the topics and writing style chosen by the user. Scribe is now used in over 100 social service agencies and all public healthcare institutions. On average, 36 minutes are saved on documentation per conversation. That time is not just a data point. It is time returned to care.
If LEGO connectors were redesigned every year, no one could build anything coherent. In Government, incompatible systems have the same effect. Previously, agencies often built separate systems for different needs, believing that every need was unique. While well-intentioned, this led to duplication and integration challenges. Citizens feel the fragmentation when information cannot be shared across systems.
So, our approach must be modular. We must provide and use common digital components, like secure logins and payment processing, built on shared standards for security and resilience. Agencies should not rebuild what already exists. They should reuse, recombine and focus on what makes their missions unique.
When the Health Sciences Authority (HSA) stepped up vaping enforcement last year, they needed a new system for their operations. HSA built on shared tools from GovTech and OGP, such as Ownself Gather, a case management system, and Plumber, which allows officers to automate manual tasks, like tracking repeat offences. By doing this, their enhanced Vaping Information System was live in just three weeks. A system built from scratch, would have taken months.
When we build faster, we enforce faster and citizens are better protected.
5.00 pm
Some services need simple, reliable components and we must resist the temptation to over-engineer. But sometimes, we do need advanced tools that can help us build better services and do so more quickly.
Mr Henry Kwek asked how GovTech is embracing AI-centric IT development and encouraging its vendors to adopt such practices. We are providing common tools, such as commercial AI coding assistants that support Government developers and vendors with tasks like code completion. We also have AI tools for GovTech officers to build and deploy functional prototypes without writing code. For both the Public Service and our contracted vendors, these tools are governed by standards for secure development, safe AI use and data protection. GovTech is also piloting agentic AI coding tools internally and plans to expand these capabilities across Government.
Our shared tools also help agencies build more inclusive and user‑centric services. As Mr Sharael Taha highlighted, essential Government services should be accessible by everyone, including persons with disabilities.
One of our most valuable shared tools is called Oobee. It proactively detects accessibility issues on Government websites and suggests fixes, like including descriptive text that can be read aloud by Assistive Technology for our visually impaired users. Oobee has scanned over 1,600 websites since 2023. It has shown us how even with good intentions, we may have blind spots.
The success of our services rests on a strong foundation of trust. Singaporeans use our digital services because they trust that these are secure and that they are dealing with legitimate Government officials. Government official impersonation scams are a serious threat because they attack that trust directly.
Mr Sharael Taha asked about our efforts to protect citizens from scams. We have already taken steps, for example, by unifying SMS messages from Government under the "gov.sg" sender ID. For citizens to easily identify and trust calls from the Government, OGP and IMDA are developing systems for agencies to make calls with numbers that start with a common prefix. Later on, we will display a recognisable caller name.
Many scammers also use local SIM cards for illicit purposes. To address this, IMDA, in consultation with the Singapore Police Force, recently implemented a limit of 10 postpaid SIM cards per person across all telcos. The Government will also apply analytics to SIM card registration data, within strict legal safeguards, to proactively detect and disrupt potential scam activities. These measures focus on identifying suspicious registration patterns.
Think about what it takes to go from building a beginner LEGO set, maybe a simple car, a few dozen pieces, designed for a young child, to a full LEGO Education SPIKE robot. It is not just more bricks. It is a different level of skill, confidence and ambition. The builder must grow with the challenge. Our public officers must grow with the challenge.
Many of our loyal and hardworking public officers have spent years building skills for their work. With the tools around them changing fast, this can feel exciting for some. For others, it feels unsettling, as if the expertise they have worked hard to develop might be overtaken before they can fully use it.
Our job is not just to offer comfort, but to build capability. To give them the confidence to use a new tool and think: I can work with this. I can ask the right questions. And I can tell whether the output is good or not.
We have already started digital training for our Cabinet Ministers and Senior Public Service Leaders, as Minister Chan Chun Sing mentioned earlier. Leaders set the conditions. When they understand the digital landscape, they can guide change confidently and ask their teams the right questions. For our broader Public Service, the goal is to ensure that no public officer feels powerless in a digital world.
MDDI will establish the Institute of Digital Government together with the Civil Service College. The Institute will equip our public officers with Digital, Data, Design and AI skills. We will focus not just on technology but on designing solutions that are citizen-centred and secure.
We also need to address the outdated systems that no longer support our needs. These are systems built early in our digitalisation journey, using technology that was prevalent at the time. Today, these systems are inflexible, expensive and difficult to integrate. They slow down policy change and hinder our ability to share information for seamless services. We have started on this. Rebuilding our systems will take time, but we are committed to this effort, because this is foundational to our digital transformation. Every modernisation effort gives us a chance to rebuild and create faster, connected systems that can better support service delivery.
I have described what we must do, to improve our Government digital services. We must understand our citizens continuously, build in a modular way and develop capabilities for complexity. We will upgrade the skills of our public officers and rebuild many of our outdated systems over time.
If we do this well, our citizens will feel the difference. Their experience with Government will feel simpler, clearer and more human, especially at moments when citizens have little bandwidth to deal with Government.
Take parents of newborns, for example. In those precious early days, what you want most is time with your baby, not multiple forms to fill. We have therefore integrated services around this life moment. Through LifeSG, parents can complete services like birth registration, apply for Baby Bonus and Shared Parental Leave, with far less paperwork. The aim is simple: fewer steps, fewer repeats, more time for families.
With AI, we can go one step further, from services that respond, to services that guide.
Take the SupportGoWhere portal, which consolidates Government schemes across 31 agencies. Imagine trying to look for support from nearly 300 options, across different life stages and needs. Our seniors and their caregivers told us that they felt overwhelmed: too much information, too many pathways. And where do they even begin? So, we redesigned the experience. The Senior Support Recommender will ask them several questions and then uses AI to recommend them the most relevant schemes. We are still refining it, but this is the direction that we want to go: a government that helps citizens find the information that they need, instead of making them hunt for it.
Delivering better services also means that our precious human Public Service officers can focus their time and their resources on supporting those who truly are not able to use technology. They can be served by our officers in a more timely manner.
Better services must also be faster, because delays are not just administrative in nature. They do affect real lives. This matters, in healthcare.
As our population ages, demand for healthcare professionals will rise. Lengthy manual registration processes to bring in nurses can become a bottleneck that delays care. And that is why we are simplifying our processes and rebuilding the Professional Registration System to streamline and automate routine checks for our healthcare professionals. We have reduced processing time for foreign nurse registrations from up to six months to 30 days. For patients and families waiting for care, this can mean earlier treatment, earlier support and less anxiety.
These examples are not exceptions to be admired. They must become more of the norm. All of us working in Government must ask ourselves: if we are responsible for a policy, is it being delivered in the simplest way possible? If we are designing a service, ask: would I accept this experience for my own family?
Our transformation will not happen overnight, but that is the standard that we will hold ourselves to.
Mr Chairman, building digital capability is not just about chasing technology. It is about raising our standard of service. Increasingly, Singaporeans compare us to the best digital experiences in their daily lives, not just to governments elsewhere. That is the standard and it is our responsibility to meet it.
The Chairman : Minister of State Rahayu Mahzam.
The Minister of State for Digital Development and Information (Ms Rahayu Mahzam) : Mr Chairman, while technology has made our lives easier and offers the promise of a better future, we must ensure our digital society remains safe and vibrant.
The Government will continue to play a strong role in protecting those who are most vulnerable to online harms. At the same time, we must also empower citizens to build the skillsets and confidence to navigate and learn in today's digital world, especially with the advent of AI. I will outline MDDI's efforts in these areas.
Let me start with online harms. Many of us have heard stories, or perhaps even know someone who is a victim of online harms. Some victims have experienced online stalking, while others have had their intimate photos abused. Often, victims and their families deal with tremendous distress and helplessness.
This is why the Online Safety (Relief and Accountability) Bill, passed in Parliament last November, is so crucial.
As part of the Online Safety (Relief and Accountability) Act, we will establish a new agency, the Online Safety Commission (OSC). The OSC will be set up by the first half of this year. It will begin by supporting victims of five highly prevalent and severe online harms: online harassment; intimate image abuse; image-based child abuse; doxxing; and online stalking. Upon assessment of a victim's report, the OSC can issue directions to disable access to harmful online content or restrict the perpetrator's online account.
Even as we set up the OSC to provide an additional avenue of support, I know that many parents are naturally worried about their children's daily digital activities. Children are amongst the most active digital users, and many parents are stretched as they juggle monitoring their children's digital usage with other commitments. In MDDI's Digital Parenting Study in 2025, over half of respondents wanted more Government support, including stronger legislation, to help them manage their children's digital activities.
Ms He Ting Ru asked about efforts to better protect children from the harms and risks associated with social media.
We have taken progressive regulatory measures to address the concerns of parents and the community. Over the past three years, we introduced two Codes of Practice for Online Safety. The Codes require designated social media services and app stores to minimise Singapore users', especially children's exposure to harmful content. The Codes also require designated social media services and designated app stores to submit annual online safety reports to IMDA. IMDA is presently assessing the annual online safety reports submitted by the designates social media services in 2025. IMDA's overall report will be published alongside designated social media services reports when ready.
5.15 pm
Today, we already do age assurance in the physical world, like how supermarkets or convenience stores check the ID of customers before selling age restricted products, such as alcohol or tobacco. From the end of this month, designated app stores will have to implement age assurance measures to prevent users who are under 18 from accessing and downloading age-inappropriate apps.
As new risks continue to emerge, online safety remains a constant challenge around the world. Some overseas jurisdictions have announced or implemented social media bans. Singapore also wants to strengthen protection for our children online and we want to do it right and take a holistic approach.
As MDDI continues to study the impact of social media bans, we plan to extend age assurance requirements to designated social media services. This would better ensure that online services are age-appropriate for users, including children. Consultations with the designated social media services are ongoing, and more details will be announced later this year.
The Government remains vigilant regarding online harms outside of app stores and social media services. Some parents have expressed concerns about harms that online video games bring, including exposure to inappropriate content, cyberbullying and screen addiction. We recognise these concerns and are studying whether safeguards on online video games are needed.
We are also mindful of other types of online services that may pose a threat to online safety. As Ms He Ting Ru pointed out, one example is the misuse of AI to generate indecent content, such as sexual content and violent content, in real time and at scale. Chatbots that are embedded in social media services present unique risks as users, including children, can access them more easily.
Ms He also raised a recent worrying trend of users using prompts on X's chatbot, Grok, to replace the clothes of adults and children with revealing attire, such as bikinis. IMDA is engaging X on the issue. We note that X has taken some measures to address the matter globally. We will continue to monitor the issue closely and will work with X to enhance online safety for Singapore users on its platform. We will not hesitate to take designated social media services to task if they have failed to comply with the SMS code. We are also studying whether safeguards for AI chatbots are needed to better protect users from the harms caused by their misuse.
While parents can look forward to stronger guardrails to protect our children from online harms, parents also play an important role in inculcating healthy digital habits in children. Parenting in the digital age is undoubtedly challenging. We frequently hear stories of children being glued to their devices during family dinners, or parents feeling shut out from their children's digital spaces. Some have even described parenting today as swimming against a relentless digital tide, while struggling to stay afloat amidst competing priorities. These concerns are real, and we want parents to know that they are not alone.
To address these concerns, MDDI has rolled out resources for parents and is strengthening efforts to make them more accessible in the community. Parents can access tips on how to guide their children's digital interactions on IMDA’s Digital for Life portal. These are tailored to different digital milestones in the parenting journey, such as a child's first device use, first social media use and first online game.
Families with young children will also be supported by digital parenting workshops and webinars. These sessions are designed to meet different needs. Some support parents of younger children, while others engage families with youths who may encounter more challenging online situations. We will continue to do more to support digital parenting and welcome suggestions on how we can improve our programmes.
Preparing for the digital world requires us to not only be safe online users, but also purposeful and discerning learners. We need to prepare our students and educators for an AI-enabled future, as Mr Henry Kwek and Ms Lee Hui Ying highlighted. Educators play a critical role in the development of essential skills for our students.
Minister of State Jasmin Lau spoke earlier on enhancing public sector digital capabilities. We are also doing more to grow educators' knowledge and understanding about technology. Last year, MDDI and MOE launched the Smart Nation Educator Fellowship, and 58 fellows attended the inaugural run. I am glad that the feedback has been positive. Many participants shared that the Smart Nation Educator Fellowship has sharpened their abilities to guide students to become thoughtful and responsible users of technology.
Mr Ezal bin Sani, a Lead Teacher for History in Jurong Secondary School, was one of the Smart Nation Educator Fellowship fellows. For a Secondary 1 History inquiry-based project on Singapore's migrant Chinese communities, Ezal’s students conducted AI-powered interviews with historical "coolies" and used ElevenLabs or Google NotebookLM to create their own AI podcasts. Initiatives like this demonstrate how powerful educational technology can be in our classrooms, with students critically assessing the information gleaned from AI tools to enhance their learning process.
This year, we are refining the fellowship to focus on the power and possibilities of AI. Through workshops and industry visits, educators will better appreciate how AI is relevant in the workplace. This can in turn support students' development of AI skills and competencies. We will update programmes in our schools to meet emerging needs, just as we have done before. As the Prime Minister shared, AI literacy is fundamental digital competency that will become even more important going forward.
Mr Sharael Taha and Mr Darryl David asked about the Government's plans to make digital skills and tools available to every child, regardless of background. We will continue to ensure that AI literacy programmes remain accessible in our schools.
IMDA is working with MOE to update the Code for Fun programme in our primary and secondary schools to integrate AI skills as core baseline capabilities for all students. We will make this available to all schools in 2027. Primary school students will learn the basics of AI, such as creating digital storybooks. Secondary school students will learn to use AI to create solutions for real-world problems. As students experiment and learn with AI, they will also learn about the risks, limitations and responsible use.
For lower-income families, IMDA's DigitalAccess@Home scheme will continue to support them with subsidised broadband and computing devices.
I thank Mr Fadli Fawzi, Ms Jessica Tan, and Mr Sharael Taha for their interest in public education initiatives to strengthen citizens' digital literacy and resilience, including for our seniors and persons with disabilities. Today, citizens can look to IMDA's Digital Skills for Life resources to pick up skills to navigate digital spaces. This includes how to use Generative AI confidently and safely, as well as how to identify AI risks, such as misinformation, scams and deepfakes.
Our libraries also provide important touchpoints for digital learning. The National Library Board's (NLB's) S.U.R.E programme, which stands for Source, Understand, Research, Evaluate, encourages Singaporeans to evaluate the credibility and reliability of information. NLB will roll out new resource packages and outreach programmes under SURE to build information literacy skills. NLB will also offer roving experiential showcases across public libraries and other public spaces. Members of the public can experience the uses and benefits of AI, and how to use GenAI safely and responsibly.
We will continue to provide targeted support for vulnerable groups. At SG Digital Community Hubs across Singapore, seniors can learn how to use digital services for daily tasks, such as booking medical appointments and mobile banking. I would like to reassure Ms Sylvia Lim that the Government will continue to adopt a "digital first, but not digital only" approach. Citizens, particularly seniors, who need in-person support will still be able to receive assistance at Government agencies' physical service touchpoints and at ServiceSG Centres.
We will also continue to collaborate with Digital for Life partners to help persons with disabilities participate meaningfully in our digital world. For instance, Guide Dogs Singapore developed a toolkit to help members of the visually impaired community learn to use low vision accessibility features on the smartphone, such as the VoiceOver function.
Through these efforts, we are building an inclusive Singapore where every citizen can benefit from our digital future. Sir, allow me to say a few words in Malay.
( In Malay ) : [ Please refer to Vernacular Speech .] Mr Chairman, mastering new skills can feel daunting sometimes, but the learning experience becomes more rewarding with support from our peers and the community. Digital technologies, including AI, have become important ingredients in our daily lives and for the jobs of tomorrow. Thus, it is important that all of us are equipped to not only have digital skills, but to use them confidently and thoughtfully. Everyone is at a different stage of their digital journey. Some are just starting to explore and experiment, while others are focused on enhancing their technical skills and putting them into practice.
To work towards making our vision of an AI-Confident Malay/Muslim community a reality, M³ has launched Langkah Digital, led by Yayasan MENDAKI. As I announced last month, Langkah Digital is designed around three key elements – Kenal, Guna, Yakin. Kenal (Knowing) helps people learn more about safe digital exploration; Guna (Use) encourages integrating technology in daily life; and Yakin (Confident) nurtures lifelong digital learning so that the community can continue to adapt independently. This step-by-step approach allows us to engage different segments of our community and provide support that matches the level of their digital journey.
Staying true to our "gotong-royong" spirit, MENDAKI will bring together the whole community, including the M³ family, Malay/Muslim organisations, the MENDAKI Professional Networks, as well as partners from the public and private sectors. A good example is Mr Luqman-nul Hakim, who works as an AI engineer and has stepped forward to support Langkah Digital.
Luqman has led sessions on AI at Al Khair Mosque for over 30 participants and facilitated an AI workshop for more than 60 participants. These programmes help cater to individuals with different abilities and interests in our community. Some introduce AI tools such as ChatGPT, while others focus on more advanced skills such as prompt engineering.
Digital champions, such as Luqman are the driving force behind Langkah Digital. They not only help to bring different groups together but also foster meaningful dialogue and collaboration around technology.
Although I just launched the programme last month, I am heartened that we have rolled out 12 AI-related workshops and events in the community so far, reaching over 400 participants. I hope that Langkah Digital will empower even more members in our community.
( In English ): Mr Chairman, as technology keeps evolving, it is not enough for us to hone our technical skills. To harness AI wisely, we need to ask the right questions and be discerning about the answers we get. Therefore, our children need to develop capabilities to read and process information effectively from young.
I am spotlighting reading because it is a skill that is increasingly at risk in a world where information is delivered at break-neck speeds, often in short-form and visual formats. Reading is essential for learning new skills. It improves our attention span, develops critical thinking skills, and builds creativity and empathy. All these are essential qualities for us to use technology for the benefit of ourselves and others. Promoting reading is therefore important in addressing the concerns raised by Members about social and intellectual degradation that might come with AI.
Ms Cassandra Lee asked how NLB is refreshing the libraries' role with families in mind. We can start with encouraging parents to cultivate good reading habits in their children from a young age. This also provides our children with a screen-free alternative in our device-heavy era.
This is why NLB will continue to partner with MOE to strengthen the library programme for schools. This includes an upcoming School Librarians of the Future Summit and webinars to empower student librarians as reading advocates and champions for information literacy. Student librarians can also broaden their learning, through volunteering, student attachments and learning journeys. NLB will do more to foster strong reading habits and plans will be shared in due course.
Our libraries also play a vital role in preserving and sharing our Singapore Stories. These stories form the bedrock of our community and ensure that our collective experiences are not lost to time.
5.30 pm
To mark Singapore's 60 years of Independence, NLB and MDDI launched a book and exhibition on The Albatross File, which documents the events, personalities and debates surrounding Singapore's journey to Independence. I thank Mr Christopher de Souza for commending the teams involved. The exhibition has resonated strongly with the public. Since December, over 130,000 have visited; 96% said they left with a deeper understanding of the path Singapore took.
To Mr Fadli Fawzi's observation that the exhibition departs from what he was taught in school, there has always been differing points of view on Singapore's separation from Malaysia. This is unsurprising, given the nature of historical accounts.
For example, British, Australian and New Zealand archives released from the early 1990s reflected the perspectives of their diplomats and governments. In 1998, full versions of the perspectives of Singapore's officials appeared. Prof Albert Lau's "A Moment of Anguish" remains the most definitive account of the separation. The first volume of founding Prime Minister Mr Lee Kuan Yew's memoir also featured a gripping account.
As stated in the editorial notes of The Albatross File: Inside Separation, nothing material to the understanding of the separation was held back. The documents published were reproduced in full, without redaction. Members of the public, including Members of Parliament, can approach the National Archives of Singapore to look at all this declassified material and form their own nuanced views.
Mr Fadli also raised the proposal for a Freedom of Information Act to automatically declassify records and release them publicly after 25 years.
The experience of some countries with similar legislation shows that most, if not all, will still have carve-outs. In fact, implementing an Freedom of Information Act could lead unintentionally to more opacity. Mr Tony Blair's 2010 memoirs record his views on the UK's Freedom of Information Act, which was enacted while he was Prime Minister. He said, after leaving office, that "The Freedom of Information Act…is a dangerous Act [because] governments need to be able to debate, discuss and decide issues in confidence. Without the confidentiality, people are inhibited and the consideration of options is limited in a way that isn't conducive to good decision making. In every system that goes down this path, what happens is that people watch what they put in writing and talk without committing to paper. It's a thoroughly bad way of analysing complex issues."
In other words, a Freedom of Information Act can hinder rather than facilitate governance because issues deemed too sensitive are simply not documented. Therefore, our starting point should be prioritising transparency that leads to good governance and an informed citizenry instead of transparency for its own sake.
We already have a mechanism for members of the public to request access to Government records for reference or research, which historians and researchers have used to nominate documents for review. The Government has made more records available to the public over time and will continue to do so.
Singapore stories not only enhance our understanding of history but enable us to reimagine our future. The SG60 Heart&Soul Experience, which was visited by two million visitors in the second half of last year, allowed Singaporeans to express their hopes and dreams for Singapore. I am glad to hear that it was well received, with most visitors giving it a rating of five out of five.
The Experience was also an example of utilising digital innovations, such as AI and immersive storytelling, to present a multisensory experience. By pairing the experience with NLB's resources, visitors had the opportunity to learn more about using AI in their daily life.
To conclude, Mr Chairman, as digital technologies become increasingly complex and sophisticated, their potential to transform our society and lives, whether for better or worse, has never been greater.
As the Malay peribahasa or proverb reminds us, "Berat sama dipikul, ringan sama dijinjing." We share responsibilities and work together to overcome challenges, be it big or small. Harnessing technology for good as well as mitigating the negative impacts of technologies require more than just technical knowledge. Ultimately, we need a whole-of-society effort that brings together the experiences and perspectives of the Government, industry, academia, civil society and citizens.
Let us work together to chart a bright and promising digital future for generations to come. [ Applause. ]
The Chairman : We have just around 25 minutes for clarifications. I will prioritise Members who have filed cuts according to the amount of time they have filed. Mr Sharael Taha.
Mr Sharael Taha : Thank you, Chairman. I thank the Minister and the political office holders for their comprehensive response. I have a few supplementary questions.
I believe the Minister did not get an opportunity to address my cut on the role of NAIC. What precisely is its role? Will it have execution authority or oversee cross-Ministry implementation or remain advisory? And how will the NAIC integrate economic strategy under MTI with digital governance under MDDI to ensure a coordinated delivery?
Also, on NAIS 2.0, what progress has been made and what did we learn from implementing NAIS 2.0 that will shape our current plans?
My last supplementary question is on trust and regulatory credibility. Minister Josephine Teo shared that Singapore is known for our trust and regulatory credibility. How can we operationalise this trust? Will we consider establishing a formal trusted AI certification regime for high-risk AI systems deployed in finance, healthcare and aviation? Maybe that could be our differentiator and competitive advantage.
The Chairman : Minister Josephine Teo.
Mrs Josephine Teo : Chairman, that is quite a few questions. I will try my best.
Perhaps to address Mr Sharael Taha's question on what we have learned from the implementation of NAIS 2.0, the Prime Minister launched NAIS 2.0 in December 2023. So, it is just a little over two years since we formally started the process. If I were to say up to now, what we are learning from it? Maybe two or three key observations.
The first, perhaps to refer to the earlier handout that MTI distributed – this is on Kampong AI – as an example. I recall that around 2023, we had visited a corner of San Francisco that was known as Cerebral Valley. It was very captivating for us because the hacker houses had such an energy about them. Papers that were being published by universities in the morning, in the evening, they already have a talk and people would come in and say, this is how I am going to use it. Then, a week later, a prototype would have already been built and venture capitalists were invited for show and tell. You could see that this generated a kind of buzz that was very enviable.
We asked ourselves: can we do that? We really did not know how comprehensive our ecosystem was at the time, that would centre its attention on AI-related activities. So, we said, let us start with Lorong AI.
It was not called Lorong AI at the start. We just decided that we just have to bring the community together. We did not have our Cerebral Valley then. We just had to see whether it could take off. It turned out to move a lot faster and grow a lot bigger than we expected. I think last year, Lorong AI hosted something like 150 events with about 4,000 people. Because of that, we have the confidence to do Kampong AI.
But what is the reflection and what is the lesson learned there? It is that you can start small but dream big. You do not always have to have a Big Bang approach. That is the first observation.
The second observation I would say is that in terms of how we have implemented ideas, actually, gaps are good opportunities. Let me explain a little bit.
After we did Lorong AI, we were also building up AI Centres of Excellence. If we did not have a good momentum with the AI Centres of Excellence, it will be hard to now try to push ahead with Champions of AI. We do not have a base to begin with.
And one of those particular Centres of Excellence was in manufacturing. When we looked at the activities being carried out at that Centre of Excellence, we realised that there are common problems that a sectorial Centre of Excellence could do. And then, it enabled us to think that perhaps in some sectors, there was a certain readiness for end-to-end transformation.
So, the national AI Missions, in a way, grew out of that effort too. Then, we asked ourselves, we now have activities in industry, we now have activities in Government, which Minister of State Jasmin Lau spoke about – AI activities. In research, things were also happening in a very positive way.
Microsoft has been in Singapore for years, but it was only last year that they set up a research entity here. Then Google DeepMind decided to open an office here.
What that tells us is that research activities are picking up pace, because AI top talents are attracted by bold ambition and a willingness to tackle the big questions, the problems that confound many people. So, we need to have that.
The gaps, being opportunities specifically apply in the talent space. We had started thinking broadly in terms of AI Creators. That would be the people in Microsoft Research, Google DeepMind. For example, in Google DeepMind, within the Singapore team, they are contributing to the development of the newest Gemini models.
So, we had thought about AI Creators, we had thought about AI Practitioners – machine learning engineers, data scientists – and we had thought about users – the broad base. As we were implementing the Centres of Excellence, we also realised that in between, you need people who both understand AI, not necessarily at the deepest level, but they have good domain expertise. There is this idea that you need "bilingual" talent – people who know enough of AI, know enough of their domain and they can marry the knowledge together.
Even in terms of how we think about the AI talent spectrum, what originally was a gap now turns out to be an opportunity for us to do something about it. That would be the second reflection.
The third relates specifically to your question about the role of NAIC.
There is a well-known saying. We use it very much to talk about social solidarity. It is an African saying. It goes, "If you want to go fast, you go alone. If you want to go far, you go together."
It turns out that it applies to AI development too. If we only confine ourselves to the effort that can be applied within the verticals – MDDI, you do this; MTI, you do this; I think we will remain effective in our own ways, but we will not get to synergise. Lorong AI can become Kampong AI because there is JTC. If JTC has not been seized with this idea, then this growth could not have quite come about.
So, the NAIC is precisely to do these things. Bring the whole together. Make the whole bigger than the sum of its parts.
Today, agencies already cooperate. It tends to be one-to-one, but what we need is many-to-many cooperation and we need the mechanism to sort things out – where to prioritise, where to apply more resources, where the exploration is heading into a dead end, what kinds of policy changes that you might need.
So, the NAIC is not just an advisory role. Our Prime Minister has many things on his plate. He advises us on a whole range of things, but to chair the National AI Council is a very strong indication that we are keen on making AI transformation real for all Singaporeans and not just something that just continues to exist as a vision. We have to turn this into reality. That is what it is intended to be.
5.45 pm
Very quickly, Mr Chairman, I think there was also a question about trust and regulatory credibility, and whether we will set up institutions to do that.
We have, in fact, taken steps. For example, quite apart from introducing model governance frameworks, we have set up a digital trust centre that is our designated AI Safety Institute. We also have the Centre of Advanced Technologies for Online Safety. These are institutions that have grown within academic organisations, but we will look for opportunities to collaborate with industry, with academia, to see how we can institutionalise this capability and turn it into an advantage for Singapore.
The Chairman : Ms Jessica Tan.
Ms Jessica Tan Soon Neo : Sir, first of all, I would like to thank Minister for talking about the AI Impact programme and democratising AI. I think it is very important, so I do want to ask a little bit of clarification around that, more from an individual and citizen standpoint. How does this pan out and how can the ordinary citizen look forward to in terms of, what do they do with regards to the National AI Impact Programme and how do they access that?
The other question I had was, related to this as well, in terms of digital literacy specific to AI. I did touch upon the point about life stages and whether in terms of the digital literacy, as we are looking at it for citizens, even for workers, how do we ensure that this is looked at from a life stage standpoint, rather than from an age or just a profile standpoint?
And another question around responsible AI was that, given that it is widely used now and, in a lot of cases, for some very critical areas in terms of decision making, how do we ensure and how will the Government look at ensuring that biases are not kind of perpetuated? And the last point on inclusivity, given that AI is going to be very much part of our lives and it is not something we can ignore, how are we going to look at the multiple languages and supporting those in terms of people who are not so proficient in English?
Mrs Josephine Teo : Mr Chairman, in terms of how citizens are going to access the National AI Impact Programme, it again will not be solely through the channels that MDDI and IMDA oversee. For one, we will work very closely with trade associations and chambers as well as professional bodies for outreach. There is very strong interest from these intermediaries to get the word out and we welcome their interest.
We already have existing programmes that serve as a useful base so, for example, the enterprise solutions will very largely be offered through the Productivity Solutions Grant scheme. The Productivity Solutions Grant is well known to many of our enterprises. The intermediaries are also familiar with it. I mentioned in my earlier speech that about 30% of the solutions are now AI-enabled. That will likely grow because of the interest and also because of the availability of the tools.
If we look at enterprises that already use some digital tools, the Member can imagine that the vendor goes to the enterprise and say, "If you add this AI tool on top of it, actually, you can do more things than you were able to previously." So, this is one way to propagate on a base that is already quite ready. It is already quite well equipped.
For individuals, I mentioned that we will start, initially, with the accountancy and legal professions and then look to extend towards fields like human resources (HR). HR, for example, easily has 50,000 members, 50,000 practitioners. So, through their own mechanisms, individuals can also be reached. So, those are the different ways in which we will propagate the access to the National AI Impact Programme.
We welcome suggestions. If anyone comes to us and say, "This is what you can do." We are very keen to understand how. I believe that our institutes of higher learning are also very much involved because they interact with the private sector and they have a lot of students who are interested to prototype solutions in order to implement them, so we will look at those as well.
The Member's question about responsible AI biases, how to make sure they are not perpetuated. The way we approach the potential risks and harms of AI is a multi-prong one. In some instances, we already have safeguards that perhaps need to be sharpened. So, for example, pornography that could be enabled by AI, child sexual exploitation material. These are already prohibited under the Penal Code. We can update the Penal Code to make it clear. We can also think of instances, like the Workplace Fairness Act. If there is a demonstration of bias in the recruitment process, then that piece of legislation can be relied upon, because it does not depend on how the unfairness, the discrimination was produced, so we can deal with it that way.
But there will be instances where existing legislation, existing laws, existing measures are not enough. So, for example, we introduced the Elections (Integrity of Online Advertising), or ELIONA, legislation. This was specifically to deal with AI-generated content that are applied to candidates in an election context, so we will continue to maintain this sort of nimble approach. Use existing laws, sharpen the laws, if and when necessary, but also do not hesitate to introduce new measures when they are needed and when it is clear what can be done about them.
So, those are the things that we will do. Inclusivity, I fully agree with you, multiple inclusivity. I fully agree with you, multiple languages — perhaps I will ask Senior Minister of State Tan to reply because he has been working on this.
The Senior Minister of State for Digital Development and Information (Mr Tan Kiat How) : Sir, to elaborate on Minister Josephine's point about inclusivity, specifically around languages.
As we know, Singapore is a multilingual society and different communities take a lot of pride in their languages, especially mother tongues. One part of the work is that MDDI looks at, specifically the "I" part of MDDI, is around information. There is a committee where we bring in academics, representatives from the media, the schools, around translation. That is the National Translation Committee, which I help to oversee.
One project that we are working on, as a very good example of how we are thinking about practical tools for AI, is about translation. And translation is not just by using any tool, any large language models (LLMs), because you need local context, local nuances, for example, not everybody understands what "chope the table" is.
So, how do you translate some of these local terms, especially local specific-context terms into different mother tongue languages? The National Translation Committee has been working with GovTech and Agency for Science, Technology and Research (A*STAR) on a number of LLMs and to fine tune a model for a local context. One of the objectives is to, hopefully, have a model which we can use by the later part of this year, to allow greater inclusivity and accessibility for those not speaking English to access our public-facing websites and other materials. And to support Public Service agencies translate many of the material, which is in English, into different mother tongue languages – Malay, Mandarin, Tamil – much more easily.
And how can we do so across the board, at-scale, across the Public Service? So, we are developing an AI tool for that and we hope to share more details in the later part of the year, but this is just one specific example how we are using AI technology. Not just "copying and pasting" what other countries are doing, but fine-tuning, adapting, customising for our local context, and do so in a way that creates value and adds that element of inclusivity for all Singaporeans. We hope to share more details later and hope to get the support of all Members.
The Chairman : We are fast approaching guillotine time, so I will take one more clarification. Mr Henry Kwek.
Mr Kwek Hian Chuan Henry : Chairman, I would like to ask MDDI on what MDDI can do to set and communicate the right expectations for our people citizenry, given the fact that we are vastly transforming the way we interact with the citizens. I applaud the MDDI for the ambition and the determination to transform the way we do e-services. Basically, in a nutshell, we are asking the Government to do more with less; do faster; do differently, not just transacting with Government, but also giving people advice, like SupportGoWhere. So, it is quite a big way of transformation of how we do things, in terms of Government service.
And we are also, instead of developing services through coding, we are asking our civil service to orchestrate AI agents to code. So, it is fairly different. With new ways of doing, there is always a risk of things not working out initially. So, how are we going to set the right expectations? Like I think Minister of State Lau has mentioned that Singaporeans have high expectation of e-services.
Maybe we could look at how Google does it, when they deliver services to all of us, they have Google Beta, Google Preview, using this to communicate to people that some services —
The Chairman : Mr Kwek.
Mr Kwek Hian Chuan Henry : — may be a trial, and how can we better communicate the expectations? Yes, I will stop here.
The Chairman : Minister Josephine Teo, please, if you could keep your response concise.
Mrs Josephine Teo : Minister of State Jasmin Lau can supplement, but maybe very quickly, on citizen expectations. We do not see it in a bad way. The Member and I use digital services all the time. And for us as citizens to want the Government to do better, in terms of digital Government services delivery, is a very natural thing. And so, we want to hold ourselves to higher standards.
But the Member's point, I think is a very useful one to keep in mind, which is that the expectations around failures, the expectations around what we may need to do to introduce more security that could come at the expense of some convenience that we have been used to, that is something that we need to socialise more people to.
In cybersecurity, we always say that there is a trade-off. When things are able to move in a very fast way, you worry about whether the security features have been properly built-in, and there may be times that we need to prioritise the security more than the convenience. So, the same for some of the services that we may have to offer, we do not know for a fact that we need that sort of thing and we will try, not to purposely, to deliberately, introduce inconvenience. But if and when we need to, we have to explain to people and do so properly. I do not know whether Minister of State Jasmin Lau wants to add on.
The Chairman : Minister of State Jasmin, you have one minute.
Ms Jasmin Lau : I thank Mr Henry Kwek for asking the clarification, because it gives us a chance to explain to the rest of Singaporeans how we try to optimise. I mentioned faster and better services just now, but it is very important that we do not view digital transformation as just one or two metrics. Our systems must be fast, but also secure. Efficient, but also resilient. Ambitious, but also trusted. And so, we cannot afford to optimise one dimension over another. It is very educational, and I hope more Singaporeans can understand that we must optimise across all dimensions – speed, security, cost and trust.
The Chairman : Thank you, Minister of State. On that note, Mr Sharael Taha, would you wish to withdraw your amendment, please?
Mr Sharael Taha : I was hoping to ask another supplementary question, but I will take your guidance, Mr Chairman.
The Chairman : No. No. No. [ Laughter. ]
6.00 pm
Mr Sharael Taha : Okay. Mr Chairman, I would like to take this opportunity to thank Minister Josephine Teo, Senior Minister of State Tan Kiat How, Minister of State Rahayu, Minister of State Jasmin Lau and the Permanent Secretary of MDDI, and the whole MDDI team for the responses and the great work that they do. I would also like to thank our GPC and fellow Members for submitting 25 cuts and with more than three hours of debate, leaving no stone unturned; and I am quite sure quite a few of us would like to ask a few more supplementary questions.
AI features prominently in this Budget. AI is not a conventional policy domain. It evolves not over decades but over months, if not weeks. Frontier models double in capacity within short cycles, agentic systems are emerging with autonomous decision-making capacity, and geopolitical contentions affect —
The Chairman: Mr Sharael Taha.
Mr Sharael Taha : — energy, data flows and supply chain. Yet, amid this ambiguity is where ambiguity and velocity —
The Chairman: Mr Sharael Taha.
Mr Sharael Taha : — MDDI must juggle many things at once, drive innovation, safeguard trust, develop the deep and broad-based skill sets, protect the vulnerable and secure national competitiveness. That is no small task, and I would like to thank our team in MDDI for doing all of that. And with that, Mr Chairman, I seek leave to withdraw my amendment.
[(proc text) Amendment, by leave, withdrawn. (proc text)]
[(proc text) The sum of $2,993,365,900 for Head Q ordered to stand part of the Main Estimates. (proc text)]
[(proc text) The sum of $119,025,200 for Head Q ordered to stand part of the Development Estimates. (proc text)]