MDDI 演講稿 · 2023-11-16
楊莉明部長在第四屆歐洲 AI 聯盟大會上的線上訪談實錄
要點
- • 新加坡的「可信 AI」三層信任：①公眾信任（公平、安全、有益；為倫理設護欄；機構問責與透明）；②政府—產業信任（開源 AI Verify，已聚集 90+ 公司含 IBM、Google、微軟、Salesforce、Red Hat）；③政府—政府信任（新歐《數碼夥伴關係》、新加坡—NIST AI Risk Management Framework 與 AI Verify 的「對映」）。
- • 國際合作必須包容——所有受 AI 影響的國家都該在場。新加坡在 ASEAN（已釋出《AI 治理與倫理指南》；明年新加坡擔任 ASEAN 數碼部長會議主席）+ 聯合國（高階 AI 諮詢機構）+ 「Digital Forum of Small States」（108 個成員）+ G7 廣島程序 + 英國 AI 安全峰會（與英國 AI 安全研究院合作）等多通道發聲。
完整譯文（繁體中文）
MDDI 英文原文譯文 · 翻譯日期: 2026-05-03
本文已從早期版本的網站遷移過來——格式可能有不一致之處。
Q1) 您能分享新加坡在「確保可信 AI」上的願景與努力嗎？
AI 安全如今是大小國家共同的關切。同時——我相信我們都看到 AI 能帶來的巨大機會。AI 顯然有變革潛力——我們相信它能被用於公共利益。對於一個根本上受勞動力規模約束的國家——新加坡——我們對 AI 的興趣——其實根植於它作為「力量倍增器」的潛力。
但我們也清楚——AI 的廣泛使用——伴隨多個領域不安感的上升——包括它如何被用於詐騙、網路攻擊、誤導與虛假資訊。構建可信生態——是「為公共利益借力 AI」的核心基礎。這意味著——把標準與治理框架放就位——確保 AI 被安全、負責任地開發與部署——並在緩釋風險的同時為創新留出空間。
我們認為——這一生態的運轉——需要 3 個層級的信任建設。
首要是「公眾信任」。要加速 AI 的廣泛採用——每個人都需要相信——這項技術是公平、安全、有益的。政府在幫助構建公眾信任、回應「AI 會擾動職場與生計」的關切上——扮演關鍵角色。這是我們必須努力的領域。我們也必須為「AI 的倫理開發與使用」設立護欄。關鍵的——我們也必須確保——組織對其 AI 使用做到「問責」與「透明」。
在這種思考的基礎上——新加坡可能是世界上最早把 AI 治理框架放就位的國家之一——我們 2019 年這樣做了——當時大概是亞洲首個。
但構建公眾信任並非政府獨自能做。產業也必須向公眾傳達——其對維護標準、遵循 AI 治理的承諾。組織同時也必須幫助員工適應 AI 採用。這些都是公眾信任的根基。
第二——政府與產業之間的信任也非常重要。把公私部門的最佳能力彙集起來——讓我們能以「安全且負責」的方式——實驗 AI 技術。今年 6 月——新加坡把治理框架與測試工具包「AI Verify」開源了。這讓模型與應用開發者、第三方測試方與研究社群——能集體地為「更可信的 AI」做貢獻。我很高興分享——已有 90 多家公司（包括 IBM、Google、微軟、Salesforce、Red Hat 等主要玩家）——加入「AI Verify Foundation」——幫助我們讓該工具包以更穩健的方式運轉。我們也歡迎歐洲公司參與 AI Verify。最近——我們也通過一個首創的「評估沙盒」——把這一工具包擴充套件到生成式 AI。同樣——我們非常歡迎歐盟成為這一倡議的一部分。
第三是——政府之間的信任。各國必須走到一起——在 AI 治理方法上做協調。AI 創新——需要共同的規則、標準與基準——才能起飛。今年 2 月——我們與歐盟簽署了《新歐數碼夥伴關係》。這或許是非約束性框架——但它涵蓋了非常全面的議題——包括數字貿易便利化、跨境資料流動、網路安全、AI 等新興技術與標準。
因此——我們希望——能探索如何在我們的 AI 治理框架之間——推動「互操作性」。它們可能來自略有不同的哲學——或並不相同——但尋找共同基礎始終有價值。這就是為什麼——新加坡與美國——最近宣佈了我們的合作。我們最近完成了一次成功的「對映」——把美國國家標準與技術研究院（NIST）的《AI 風險管理框架》——與新加坡的 AI Verify——做了對照。所以——我們相信——國際合作的基礎已開始建立——這將促進「負責任且安全的創新」——為我們的人民與企業帶來切實的好處。
Q2) 新加坡如何為「全球可信 AI」的國際合作做貢獻？
我先後退一步說——一談到國際合作——我們必須盡力做到「包容」。這意味著——所有「人民與企業會受到 AI 廣泛開發與部署影響」的國家——我們都希望成為這場對話的一部分——不希望被遺漏。
我們也處在 AI 發展的關鍵節點——「作為一個國際社群協作」——會讓我們更有機會確保「人人都能以安全、可靠的方式充分收穫 AI 收益」。所以我們必須繼續推動「多利益相關方交流」——把多元視角帶到桌前——不能由少數人「替我們所有人發言」。
因此——我們希望繼續鼓勵同行——以「包容」的方式——確保較小與發展中國家——在塑造「健康的國際可信 AI 環境」時——有發言權。
也分享——在我們所在的 ASEAN 區域——成員國處在 AI 發展與數字訪問的不同階段。但我們仍能合作開發出《ASEAN AI 治理與倫理指南》——給我們一個區域級的「基線」來指引前進。明年——新加坡擔任「ASEAN 數碼部長會議」主席時——我們將繼續在這一基礎上做貢獻並擴充套件。
若把合作擴到聯合國層面——我們當然歡迎「AI 高階諮詢機構」（high-level advisory body on AI）的成立——它將與多邊層面的多項數字倡議互補——並把跨地區、多元的群體聚到一起。
在成員國之間——新加坡通過「Digital Forum of Small States」（共 108 個成員）——支援我們的全球數字目標。這是一個我們分享經驗、彼此支援構建能力的社群——好讓每個國家都能在全球 AI 對話中有意義地參與。這些努力將與 G7 廣島程序等其他全球對話互補。
本月早些時候——我出席了在英國舉辦的「AI 安全峰會」——它彙集了不同利益相關方討論 AI 安全（尤其涉及基礎模型）的重要議題。參與者跨越政府、產業專家、研究社群與公民社會。峰會上宣佈——新加坡將與英國「AI 安全研究院」合作——共同推進 AI 測試方面的能力與專長。最終——我們希望共同創造可信 AI——能切實惠及經濟與社會。
演講實錄的 PDF 版本
英文原文
MDDI 官網原始記錄 · 抓取日期: 2026-05-02
This article has been migrated from an earlier version of the site and may display formatting inconsistencies.
Q1) Could you share the vision and efforts of Singapore to ensure trustworthy AI in your country?
AI safety is now an issue of concern for all countries, both large and small. At the same time, I think we all see the tremendous opportunities that AI can bring. AI certainly has transformative potential, and we believe can be harnessed for the public good. And for a country like Singapore that is fundamentally constrained by the size of our workforce, our interest in AI is really grounded in its potential as a force multiplier.
However, we are also aware that the use of AI on a widespread basis is accompanied by a growing sense of unease in several areas, including how they can be used for scams, cyberattacks, and for misinformation and disinformation. Building a trusted ecosystem is the core foundation for harnessing AI for the public good. And this means putting in place standards and governance frameworks to ensure that AI is developed and deployed safely and responsibly, and that we can mitigate risks, while maximising the room for innovation.
We believe there are three levels of trust building that are necessary for this ecosystem to operate and function.
First and foremost is public trust. To accelerate the widespread adoption of AI, everyone needs to feel that the technology is fair, safe and beneficial. Governments play an essential role in helping to build the public trust, to address concerns that AI will bring disruption to workplaces and livelihoods. That is one area that we must work on. We must also set up guardrails for the ethical development and use of AI. And critically, we must also ensure that organisations are accountable and transparent about the use of AI.
As a result of this thinking, Singapore was probably one of the first countries in the world to put in place an AI governance framework. We did this in 2019. At the time, we were probably the first in Asia.
Now, building public trust is not just for the government alone. Industry must also communicate to the public its commitment to uphold standards and adhere to AI governance. Organisations must at the same time help their workers to adapt to AI adoption. So, these are the foundations of public trust.
Second, the trust between government and industry is also very important. Harnessing the best of public and private sector capabilities allows us to experiment with AI technologies in a safe and responsible way. In June this year, Singapore open sourced our governance framework and testing toolkit called AI Verify. This allows us to involve model and app developers, third-party testers and the research community to collectively contribute to more trustworthy AI. I am glad to share that over 90 companies, including major players like IBM, Google, Microsoft, Salesforce, and Red Hat, are now part of the AI Verify Foundation to help us make this toolkit work in a more robust manner. We also welcome European companies to participate in AI Verify. We also recently extended this toolkit to generative AI through a first of its kind evaluation sandbox. Again, we very much welcome the EU to be part of this initiative.
The third area is the trust that we will need between governments. There is certainly a need for countries to come together to harmonise the approaches around AI governance. AI innovation needs common rules, standards and benchmarks to take off. With the EU, we have signed the EU- Singapore digital partnership in February this year. This may be a non-binding framework, but it covers a very comprehensive range of issues including digital trade facilitation, cross border data flows, cybersecurity, and emerging technologies such as AI and standards.
So, we hope that we can explore ways to promote interoperability between our AI governance frameworks. They may come from a slightly different philosophy, and they may not be identical, but there will always be value in us trying to find common ground. And this is the reason why Singapore in the US announced our recent collaboration. We recently went through a successful mapping between the US National Institute of Standards and Technologies’ AI Risk Management Framework and Singapore's AI Verify. So, we are confident that the foundation for international collaboration has already begun to be built. And it will help to promote innovation that is responsible and safe and can lead to tangible benefits for both our people as well as our businesses.
Q2) How does Singapore see itself contributing to international cooperation for trustworthy AI globally?
I will take a step back to first suggest that when it comes to international cooperation, we really must try our best to be inclusive. And by that, we mean that all the countries whose populations whose people and businesses will be impacted by the widespread development and deployment of AI. We all want to be part of this conversation, and we don't want to be left out.
We are also at a critical point in AI development, where working together as one international community will give us a better chance of ensuring that everyone can fully reap the benefits of AI in a safe and secure manner. We must therefore continue to promote multi-stakeholder exchanges to bring diverse perspectives to the table. It cannot be a few speaking for all of us.
And so, we want to continue to encourage our colleagues to be inclusive in ensuring that the smaller as well as developing countries have a voice in shaping a healthy international environment for the trustworthy use of AI.
Now, let me also share that within the region in ASEAN, the member states are of course in different stages of AI development and digital access. Yet, we have found it possible to work together to develop an ASEAN guide on AI governance and ethics. This allows us as a region to have a baseline to guide our progress. We will continue to contribute as well as to build on this work when Singapore chairs the ASEAN Digital Ministers’ Meeting next year.
If we then broaden this cooperation to the UN level, we certainly welcome the setting up of the high-level advisory body on Artificial Intelligence. It will complement the various ongoing Digital Initiatives at the multilateral level and bring together an important cross-regional and diverse group.
Among the member states, Singapore is supporting our global digital objectives through the Digital Forum of Small States, which comprises 108 members. This is a community through which we share experiences and support one another in building capacity, so that everyone can participate meaningfully in the global AI conversation. These efforts will complement other ongoing global conversations on AI such as the G7 Hiroshima process.
And earlier this month, I was at the AI Safety Summit that was held in the United Kingdom, which brought together various stakeholders to discuss the important issue of AI safety, especially with foundation models. Participants cut across governments, industry experts, as well as the research community and civil society. It was announced at the Summit that Singapore will be working with the UK AI Safety Institute to collectively advance capabilities and expertise in AI testing. Ultimately, we want to co-create trustworthy AI that can effectively benefit the economy as well as society.
PDF version of the transcript