A US 501(c)(3) nonprofit dedicated to advancing international cooperation to reduce extreme AI risks, best known for running the International Dialogues on AI Safety (IDAIS) series that convenes leading scientists from around the world.
A US 501(c)(3) nonprofit dedicated to advancing international cooperation to reduce extreme AI risks, best known for running the International Dialogues on AI Safety (IDAIS) series that convenes leading scientists from around the world.
People
Updated 05/18/26Executive Director
Events Associate
Programme Manager
Affiliate
Research Fellow
Research Fellow
Head of Operations
Senior AI Policy Researcher
Funding Details
Updated 05/18/26- Annual Budget
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
Org Details
Updated 05/18/26The Safe AI Forum (SAIF) is a US 501(c)(3) nonprofit dedicated to advancing international cooperation to reduce extreme AI risks. The organization was co-founded by Fynn Heide and Conor McGurk in late 2023 and was initially incubated as a fiscally sponsored project of FAR.AI. SAIF transitioned to independent operations in May 2025 after obtaining its own 501(c)(3) designation, with the move to independence having been planned from the outset. SAIF's flagship program is the International Dialogues on AI Safety (IDAIS), a series of high-level convenings that brings together senior computer scientists, Turing Award winners, and AI governance experts from around the world to build consensus on frontier AI risks and governance interventions. Four major dialogues have been held to date: IDAIS-Oxford at Ditchley Park in October 2023, IDAIS-Beijing in March 2024, IDAIS-Venice in September 2024, and IDAIS-Shanghai in July 2025. A fifth dialogue, IDAIS-London, is planned for early 2026. Each dialogue produces a consensus statement signed by participants, outlining shared goals and policy recommendations. Notable participants have included Turing Award winners Yoshua Bengio, Geoffrey Hinton, and Andrew Yao, as well as Stuart Russell, Max Tegmark, Gillian Hadfield, and many other leading figures. A distinctive feature of IDAIS is its success in bridging Western and Chinese AI research communities. The dialogues regularly include prominent scientists from both sides, fostering rare direct collaboration on AI safety between researchers who might otherwise have limited channels for engagement. IDAIS-Shanghai in July 2025 marked Geoffrey Hinton's first trip to China. Beyond IDAIS, SAIF conducts research on AI governance and safety. The team has published the IDAIS Policy Guide (February 2025), which connects dialogue consensus recommendations to actionable policy guidance for policymakers, philanthropists, companies, and researchers. SAIF researchers have also published work on AI alignment and deception, including a primer available in both English and Chinese, a position paper on preparing for AI systems that deceive developers, research on areas of technical AI safety where geopolitical rivals could cooperate, and a study on promising topics for US-China dialogues on AI safety and governance. SAIF also provides advisory services to other nonprofits aligned with its mission and runs workshops on key topics. In 2025, SAIF co-hosted an International Workshop on AI Deception Risks and Governance with Fudan University. The organization participated in the 2025 Singapore Conference on AI (SCAI) and contributed to the Singapore Consensus on Global AI Safety Research Priorities. SAIF is running a research stream at MATS (ML Alignment Theory Scholars) for Summer 2026, focusing on international coordination to reduce frontier AI risks with an emphasis on US-China relations. The organization operates with a team of nine staff based across the UK and US, led by Executive Director Fynn Heide. Co-founder Conor McGurk departed in August 2025 to join Open Philanthropy. Adam Gleave, CEO of FAR.AI, serves on SAIF's board of directors in a personal capacity to provide continuity and strategic guidance.
Theory of Change
Updated 05/18/26SAIF operates on the theory that reducing catastrophic and existential risks from advanced AI requires international scientific consensus and coordination, particularly between major AI-developing nations. By convening the world's most influential AI scientists and governance experts in structured dialogues, SAIF builds shared understanding of frontier AI risks and produces consensus statements that establish clear norms and red lines for AI development. These consensus positions then serve as authoritative references for policymakers, companies, and researchers making decisions about AI governance and safety. The organization specifically targets the US-China divide in AI governance, creating rare forums where leading scientists from both communities can engage directly, under the premise that AI safety challenges are global in nature and cannot be solved by any single nation acting alone. SAIF further amplifies impact through research that translates dialogue consensus into concrete policy recommendations and through advisory work that supports other organizations in the AI safety ecosystem.
Grants Received
Updated 05/18/26Projects
Updated 05/18/26
A series of high-level dialogues hosted by the Safe AI Forum that brings together senior computer scientists and AI governance experts from around the world to build international collaboration on AI risks.
Discussion
Key risk: IDAIS consensus outputs may not translate into concrete, binding policy or corporate practice—particularly given volatile US–China relations—so the counterfactual impact of costly convenings could be limited.
Case for funding: SAIF’s IDAIS series has a demonstrated ability to convene top-tier Western and Chinese AI scientists (e.g., Bengio, Hinton, Yao) to produce cross-border consensus statements and policy guides, a rare leverage point for establishing shared norms and reducing arms-race dynamics in frontier AI.