Cooperative AI Foundation
The Cooperative AI Foundation supports research aimed at ensuring that advanced AI systems are capable of cooperating safely and beneficially with humans and with each other. CAIF takes a grantmaking-first approach, funding academic researchers working on topics such as multi-agent safety, opponent shaping, cooperative game theory, and the evaluation of cooperation-relevant AI capabilities and propensities. Beyond grants, it runs a PhD fellowship program, organizes workshops at major ML conferences, hosts seminar series, and runs annual summer schools. The foundation prioritizes research that is important, neglected, and tractable, and emphasizes differential progress — advancing cooperative capabilities while limiting dual-use capabilities such as deception or manipulation.
Funding Details
- Annual Budget
- $1,731,887
- Monthly Burn Rate
- $144,324
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $15,000,000
- Fiscal Sponsor
- -
Theory of Change
CAIF believes that cooperation failures — between AI systems, between AI and humans, and between human actors mediated by AI — represent a major and underappreciated source of catastrophic risk from advanced AI. Multi-agent interactions can produce qualitatively different failure modes than single-agent AI safety problems. By funding academic research on cooperative AI now, while the field is nascent and neglected, CAIF aims to establish scientific foundations, evaluation methods, and practical techniques that AI developers, policymakers, and oversight bodies can draw on as systems become more powerful. The foundation emphasizes differential progress: pushing forward research on beneficial cooperative capabilities and governance mechanisms while explicitly avoiding research that advances coercive or deceptive dual-use capabilities. It also invests in community-building — fellowships, summer schools, workshops, and seminar series — to grow the pipeline of researchers working on these problems and embed cooperation science into mainstream AI development.
Grants Received
No grants recorded.
Projects
An annual intensive summer school run by the Cooperative AI Foundation that trains students and early-career researchers in the emerging field of cooperative AI.
People
No linked people.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Apr 2, 2026, 10:10 PM UTC
- Created
- Mar 19, 2026, 10:31 PM UTC