Singapore Consensus on Global AI Safety Research Priorities
Issued by
Government of Singapore and International Summit Participants
The Singapore Consensus is an international consensus document establishing a coordinated agenda for AI safety research priorities across jurisdictions and institutions. It emerged from a multilateral summit convened by the Singapore Government and reflects agreement among participating governments and organizations on where collaborative safety research efforts should be directed. The document does not impose binding legal obligations but provides a shared reference framework for national AI safety programs and research funding bodies.
Applies To
Overview
Published in May 2025 following an international AI safety summit hosted by Singapore, the Singapore Consensus outlines agreed global priorities for AI safety research, intended to align efforts across governments, academic institutions, and industry participants. The document functions as a coordination instrument, analogous in structure and intent to the Bletchley Declaration of 2023, identifying thematic research areas deemed most critical to mitigating risks from advanced AI systems. Participating jurisdictions and organizations are expected to reference the consensus when designing national safety research programs, funding allocations, and international cooperation agreements. The framework does not establish a binding compliance regime, enforcement mechanism, or penalty structure, and adherence is voluntary. Its practical effect is to shape the priorities of national AI safety institutes, multilateral research partnerships, and standards bodies that subsequently translate research findings into technical standards or regulatory guidance. The Singapore Government, as convening authority, holds the secretariat function, though ongoing governance arrangements for the consensus had not been fully detailed at the time of publication.
Key Requirements
- •Participating governments and institutions are expected to align national AI safety research agendas with the consensus priority areas identified in the document
- •Signatories and endorsing organizations are encouraged to report on research activities and progress against the identified priority themes through existing international coordination channels
- •Research funding bodies are invited to reference the consensus when evaluating proposals for AI safety research grants and programs
- •International collaboration agreements between AI safety institutes should be informed by the shared priorities established in the consensus
- •No binding compliance deadlines, financial thresholds, or penalty provisions are established; the framework operates on a voluntary coordination basis
What Your Organization Must Do
- →Assign your Chief AI Officer or Head of AI Governance to review the Singapore Consensus priority areas and produce a gap analysis comparing current internal AI safety research commitments against the consensus themes by Q3 2025.
- →Update your AI safety research roadmap and internal funding allocation criteria to explicitly reference the Singapore Consensus priority areas, ensuring alignment is documented before the next annual budget cycle.
- →If your organization participates in international AI safety partnerships or multilateral research agreements, instruct your legal and policy teams to incorporate the consensus priorities as a reference standard in new cooperation agreements negotiated from May 2025 onward.
- →Prepare a brief voluntary progress report on your AI safety research activities mapped to the consensus themes, to be submitted through relevant international coordination channels such as AI safety institute networks or multilateral forums, timed to align with any reporting cycles established by participating governments.
- →Direct your government affairs and regulatory monitoring teams to track how national AI safety institutes and standards bodies in your operating jurisdictions translate the consensus into binding technical standards or funding requirements, flagging any developments that create downstream compliance obligations within 60 days of publication.
- →Engage your research grant and procurement teams to update evaluation criteria for AI safety vendor and academic partnerships so that proposals demonstrating alignment with the Singapore Consensus priority areas receive documented consideration during selection processes.
Playbook Guidance
Step-by-step implementation guidance for compliance teams.
Frequently Asked Questions
- Is the Singapore Consensus legally binding on companies or governments?
- No. The Singapore Consensus is a voluntary coordination framework with no binding legal obligations, enforcement mechanisms, or penalty provisions. Participating governments and organizations are encouraged to align their AI safety research agendas with its priorities, but adherence is entirely voluntary.
- How does the Singapore Consensus compare to the Bletchley Declaration?
- Both are non-binding multilateral instruments designed to coordinate AI safety efforts across jurisdictions. The Bletchley Declaration of 2023 focused on frontier AI risks and established early international dialogue, while the Singapore Consensus advances that agenda by setting specific shared research priorities intended to guide national programs and funding bodies.
- Which organizations are expected to act on the Singapore Consensus priorities?
- National AI safety institutes, research funding bodies, academic institutions, government AI programs, and large AI developers participating in international safety partnerships are the primary intended audiences. The framework specifically invites these bodies to reference the consensus when designing programs, allocating grants, and negotiating cooperation agreements.
- Does the Singapore Consensus create any downstream compliance obligations for private companies?
- Not directly. However, national AI safety institutes and standards bodies in participating jurisdictions may translate the consensus priorities into binding technical standards or funding requirements over time. Compliance teams should monitor how their operating jurisdictions incorporate the consensus into domestic regulatory or procurement frameworks.
- What reporting obligations does the Singapore Consensus place on endorsing organizations?
- Reporting is voluntary and informal. Endorsing organizations are encouraged to share progress on AI safety research activities mapped to the consensus themes through existing international coordination channels, such as AI safety institute networks or multilateral forums, with no prescribed format, deadline, or verification process.
- Which government holds secretariat responsibility for the Singapore Consensus?
- The Government of Singapore holds the secretariat function as the convening authority for the summit that produced the document. Ongoing governance arrangements, including any future review or amendment process, had not been fully detailed at the time of the consensus publication in May 2025.