Policy / Project Profile
Singapore Consensus on Global AI Safety Research Priorities
A living document on global AI safety research priorities produced by SCAI: ISE 2025: 100+ participants from 11 countries, with continued updates through ISE 2026.
- Category
- International Collaboration
- Published / Updated
- 2025-05
- Issuing body
- Singapore Conference on AI (SCAI) / Singapore AI Safety Institute
- Linked levers
- Lever 6: Diplomacy
- Timeline years
- 2025, 2026
Strategic Context
Connected to 1 national AI lever.
Appears in the timeline around 2025, 2026.
Detailed Notes
The Singapore Consensus is the outcome document of the 2025 Singapore Conference on AI (SCAI): International Scientific Exchange on AI Safety. The official account: on 26 April 2025, 100+ participants from academia, industry, and government, spanning 11 countries, collectively identified and demonstrated consensus around high-priority technical AI safety research areas.
It is not an intergovernmental agreement "signed by 11 countries." The more accurate description is: a living document contributed to by 100+ participants from 11 countries, continuing to welcome feedback from the global research community.
The document builds on the International AI Safety Report chaired by Yoshua Bengio and backed by 33 governments. Its goal is to provide collaborative technical priorities for global AI safety research. ISE 2026, held from 17 to 19 May 2026, reconvened global experts to update the Singapore Consensus in light of agentic deployments, AI misuse, capability advances, and new safety research.
This remains one of Singapore’s most important international governance levers. Its leverage is not the number of signatures; it is the ability to put academia, industry, and government onto the same technical problem list.