Every AI Policy Ever · open in the interactive register

Published (non-binding, current)Non-binding · Report (government-commissioned)

The Singapore Consensus on Global AI Safety Research Priorities

Record SG-07

Issued by
Government of Singapore (IMDA / Ministry of Digital Development and Information), from the Singapore Conference on AI: International Scientific Exchange on AI Safety (SCAI: ISE)
Jurisdiction
Singapore
Level · region
National · South & Southeast Asia
Published / adopted
2025-05-08
Language
English

Summary

Outcome document of the 2025 Singapore Conference on AI: International Scientific Exchange on AI Safety, drafted by an expert committee and revised with feedback from more than 100 participants from 11 countries, from academia, AI developers, governments and civil society. Focused on general-purpose AI and building on the International AI Safety Report, it uses a defence-in-depth model to group technical AI safety research priorities into three areas: risk assessment, including metrology and enablers for third-party audits; development of trustworthy, reliable and secure systems (specification, design, verification); and control, covering monitoring, intervention and societal resilience.

Safety topics

research prioritiesgeneral-purpose / frontier AIevaluation & testingcatastrophic riskinternational cooperation

Open question

the checker found something possibly newer: {'title': 'The 2026 Singapore Consensus on Global AI Safety Research Priorities (July 2026)', 'url': 'https://www.imda.gov.sg/assets/22dddecc-661b-4fee-9634-ea5549fa8151.pdf', 'note': 'second edition; stored in this run; builds on and updates the 2025 repo Issue #158

The record shows its last checked values until a maintainer decides; the decision is recorded in its history.

Status notes

The exchange was held on 26 Apr 2025. An Expert Planning Committee (Dawn Song, Lan Xue, Luke Ong, Max Tegmark, Stuart Russell, Tegan Maharaj, Ya-Qin Zhang and Yoshua Bengio) drafted the Consensus for consultation, and it was revised after several rounds of participant feedback. It is dated 8 May 2025 and is a living document that continues to welcome views from the global research community. A second edition, The 2026 Singapore Consensus on Global AI Safety Research Priorities (July 2026), came out of the 2nd International Scientific Exchange on AI Safety in May 2026: it builds on the 2025 report, makes societal resilience a fourth pillar and adds a companion report on agentic risk management.

Sources

Open the official source https://file.go.gov.sg/sg-consensus-ai-safety.pdf

Secondary source: https://www.scai.gov.sg/2025/scai2025-report/

Local copy

Archived. Last attempt 2026-10-18 03:22:44 UTC

Official source (third-party copy, not published) archive/SG-07/2026-10-18/official.pdf
Retrieved 2026-10-18 03:22:48 UTC
SHA-256 9f4db45907903b60196440a2878b9afd5c5086b6932c1454ea873c45118aefbb
Secondary source (third-party copy, not published) archive/SG-07/2026-10-18/secondary.html
Retrieved 2026-10-18 03:22:52 UTC
SHA-256 f0ed6e0f115f083b3f11d22aa0a6abd29cc95baa48b3e9ac23589c56c16e712d

Verification

A
The instrument's own official text, or an official register, was stored and confirms the title, date and status. checked 2026-10-18 · official-url-fetched
Recheck of 18 Oct 2026. Stored and read in this run: the Consensus itself (PDF behind go.gov.sg/sg-consensus-ai-safety, on file.go.gov.sg, Government of Singapore), whose cover gives the title and 8 May 2025 and whose introduction gives the conference, the participants, the scope and the three areas; the Singapore Conference on AI's page for the report (scai.gov.sg, last updated 10 Sep 2026), which calls it SCAI 2025's outcome document and a living document; and IMDA's PDF of The 2026 Singapore Consensus (imda.gov.sg, July 2026). Also stored: IMDA's press release of 8 May 2025 (imda.gov.sg, copy of 11 Oct 2026), whose text layer holds only its heading and dateline because the body is inside a script payload the text layer does not include, and the Singapore AISI's page on the Consensus (sgaisi.sg, copy of 11 Oct 2026; on 18 Oct 2026 the site returned only a JavaScript shell). The Consensus PDF and the SCAI page are proposed as official_url and secondary_url in place of the press release and the sgaisi.sg page. No stored text names the Ministry of Digital Development and Information as an issuer; issuing_body is left as in version 1 (open question). Version 1's notes that the conference ran alongside ICLR 2025, that the Consensus was presented at the Asia Tech x Singapore Summit and that it is on arXiv rest on no stored file and were removed.
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 9 field(s) supported.

History

  1. 2026-10-18notes: Product of SCAI: ISE, held on 26 Apr 2025 in Singapore alongside ICLR 2025. Publ → The exchange was held on 26 Apr 2025. An Expert Planning Committee (Dawn Song, L — drafting process and the 2026 edition added from stored files; version 1's ICLR, ATxSummit and arXiv remarks and its characterisation as 'not a formal policy instrument' rest on no stored file and were removed2026-10-18-recheck-south-asia-1 · Claude Code (claude-opus-5-5)
  2. 2026-10-18summary: A consensus document from about 100 AI scientists, industry and government parti → Outcome document of the 2025 Singapore Conference on AI: International Scientifi — rewritten from the Consensus itself: 'more than 100' participants (version 1: 'about 100'), their backgrounds, the general-purpose AI scope and the three areas as the document describes them2026-10-18-recheck-south-asia-1 · Claude Code (claude-opus-5-5)
  3. 2026-10-18secondary_url: https://sgaisi.sg/resources/the-singapore-consensus-on-global-ai-safety-research → https://www.scai.gov.sg/2025/scai2025-report/ — the Singapore Conference on AI's report page (scai.gov.sg), stored in this run, in place of the sgaisi.sg page, which returned a JavaScript shell to the archiver on 18 Oct 20262026-10-18-recheck-south-asia-1 · Claude Code (claude-opus-5-5)
  4. 2026-10-18official_url: https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-r → https://file.go.gov.sg/sg-consensus-ai-safety.pdf — the Consensus's own text (PDF on file.go.gov.sg, linked from the scai.gov.sg report page), stored in this run, in place of IMDA's press release, whose text layer is empty2026-10-18-recheck-south-asia-1 · Claude Code (claude-opus-5-5)

Related records

Singapore AISI (Digital Trust Centre) Singapore · Published (non-binding, current)

International AI Safety Report 2025 International AI Safety Report (AI Safety Summit process) · Superseded

Citation

Government of Singapore (IMDA / Ministry of Digital Development and Information), from the Singapore Conference on AI: International Scientific Exchange on AI Safety (SCAI: ISE). The Singapore Consensus on Global AI Safety Research Priorities. 2025-05-08. https://file.go.gov.sg/sg-consensus-ai-safety.pdf. Status: Published (non-binding, current); verified 2026-10-18 (level A). Local copy archive/SG-07/2026-10-18/official.pdf retrieved 2026-10-18 03:22:48 UTC, SHA-256 9f4db45907903b60196440a2878b9afd5c5086b6932c1454ea873c45118aefbb. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record SG-07.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.