Cooperative Ai Foundation

Quick Stats

• Annual Giving: £660,000 (2024, representing less than 50% of budgeted capacity)
• Success Rate: Not publicly disclosed (3 grants funded in 2024 from full applicant pool)
• Decision Time: Approximately 5 months (application to final decision)
• Grant Range: £10,000 - £385,000
• Geographic Focus: International

Contact Details

Website: www.cooperativeai.org

Email: operations@cooperativeaifoundation.org

Phone: 01600 717615

Registered Address: Courtenay House, Pynes Hill, Exeter, EX2 5AZ

Application Portal: SurveyMonkey Apply (cooperativeai.smapply.io)

Overview

The Cooperative AI Foundation (CAIF) was established with a $15 million philanthropic commitment from Macroscopic Ventures (formerly the Center on Emerging Risk Research). Registered in England and Wales in 2021 (Charity Number 1201294), CAIF's mission is to support research that will improve the cooperative intelligence of advanced AI for the benefit of all. The Foundation uses a framework based on importance, neglectedness, and tractability to guide grantmaking decisions, with a focus on counterfactual and long-term improvements over the next 10-20 years. Beyond grantmaking, CAIF organizes workshops at major machine learning conferences, hosts online seminar series, offers PhD fellowships, and administers research prizes and tournaments. The Foundation takes a “differential progress” approach, prioritizing research that advances cooperative capabilities over dual-use capabilities that could enable coercion or manipulation.

Funding Priorities

Grant Programs

Standard Research Grants: £10,000 - £385,000 (median: £150,000)

• Duration: Up to 2 years
• No upper limit on funding amount
• Covers personnel, materials, travel, and publication expenses
• Maximum 10% indirect costs
• Two-stage application process (pre-proposal, then full proposal if invited)
• Two grant rounds annually

Early-Career Track: Up to £100,000

• Duration: Up to 12 months
• For researchers within 2-3 years of PhD completion (or similar career stage)
• Projects primarily carried out by a single individual
• Assessed on both project merit and career development potential

Priority Areas

The Foundation evaluates proposals based on three key criteria: importance (contribution to cooperative intelligence of advanced AI systems), neglectedness (unlikely to be done otherwise), and tractability (conducive to making progress).

2025 High-Priority Research Areas:

• Theoretical identification and definition of cooperation-relevant capabilities
• Rigorous arguments on desirability of such capabilities
• Empirical methods for evaluating/measuring capabilities in frontier AI systems

• Research on AI systems' tendencies toward cooperative behavior

• Mechanisms to promote cooperative behavior in multi-agent systems

• How AI can help humans cooperate more effectively

• Understanding and preventing unwanted coordination between AI systems

• Tracking and understanding multi-agent interactions

• How information differences affect cooperation

• Security considerations in multi-agent AI systems

What They Don't Fund

• Projects with budgets under £10,000
• Educational or advocacy projects (must be research-focused)
• Projects not relevant to advanced AI systems or future systems
• Research not focused on multi-agent/cooperation problems
• Applications from CAIF staff and trustees

Ready to write a winning application for Cooperative Ai Foundation?

Our AI helps you craft proposals that match their exact priorities. Save 10+ hours and increase your success rate.

Get Free Beta Access

Governance and Leadership

Board of Trustees:

• Prof. Gillian Hadfield (Chair) - Legal scholar and AI researcher
• Audrey Tang - Taiwan's Cyber Ambassador and former Digital Minister
• Allan Dafoe - Works on safety and governance of frontier AI, believes “the governance of AI is the most important issue of the century”
• Eric Horvitz - Distinguished AI researcher
• Jesse Clifton - AI safety researcher

Leadership Team:

• David Norman - Managing Director
• Lewis Hammond - Research Director (previously Acting Executive Director)

The Foundation is guided by its framework: “CAIF is guided by the importance, neglectedness, and tractability of potential activities, while maintaining an awareness of the value of curiosity and serendipity for great science.” They prioritize “counterfactual and long-term improvements over the next 10-20 years (or longer) that would have been unlikely without our support.”

Application Process and Timeline

How to Apply

Two-Stage Process (introduced 2025):

• Invitation does not guarantee funding
• Interactive feedback may be provided during proposal development

Application Requirements:

• Select one or more predefined research areas that align with your proposal
• Demonstrate alignment with CAIF's mission and priorities
• Budget must be detailed and justified
• Projects must focus on AI development for multi-agent/cooperation problems
• Results must be relevant for advanced AI systems, including future systems

Decision Timeline

• Application Period: Typically opens in late year/early year
• Decision Timeline: Approximately 5 months from application to final decision
• Grant Rounds: Two rounds per year (reduced from four in 2024)
• Notification: Decisions communicated directly to applicants

Success Rates

Success rates are not publicly disclosed. In 2024, CAIF funded 3 proposals totaling approximately £660,000, with 2 additional projects (£170,000) pending revisions. This represented less than 50% of budgeted funds, indicating the Foundation struggled to find sufficient high-quality, in-scope proposals rather than facing overwhelming competition.

Reapplication Policy

No explicit restrictions on reapplication. The Foundation has emphasized the importance of ensuring proposals align with their updated, more specific scope and priorities. For 2025, they have made the scope much more precisely defined to help applicants assess fit before applying.

Application Success Factors

Key Alignment Factors

1. Scope Alignment is Critical: The Foundation has acknowledged that many 2024 proposals were out of scope. Applicants must carefully review the specific research area descriptions and ensure their work fits within the defined priorities.

1. Focus on Differential Progress: CAIF prioritizes research that advances cooperative capabilities relative to dual-use capabilities. As they state: “CAIF wants to support research that leads to significant progress on cooperative capabilities – capabilities that lead to increases in social welfare in a wide range of environments – relative to progress on capabilities that are dual-use (e.g., useful for deception, manipulation, disempowering other agents).”

1. Demonstrate Counterfactual Impact: Show that your research is unlikely to happen without CAIF support. The Foundation explicitly looks for work that addresses neglected areas.

1. Long-Term Relevance: Projects should contribute to improvements over a 10-20 year horizon, not just immediate applications.

Examples of Funded Projects

• FOCAL at Carnegie Mellon University: Establishing a research lab focused on decision and game theory for increasing cooperation abilities of advanced machine agents
• Foerster Lab for AI Research (FLAIR) at University of Oxford: Machine learning research focused on cooperation
• Cooperative Norms and Communication: Research on how cooperative norms emerge from self-interested AI agents, including development of GovSim benchmark
• Coercion and Cooperation: Addressing dual-use capabilities underlying coercion while recognizing their necessity for cooperation
• Opponent Shaping: Evaluating methods for making decisions that guide opponents' learning toward better outcomes in mixed-motive games
• Multi-Agent Reinforcement Learning Contest: Sponsored competition based on Melting Pot evaluation suite (672 submissions, 117 teams, $10,000 prize pool)

What Strengthens Applications

• Formal research training and degrees (especially PhD) strengthen proposals but are not required
• Institutional affiliation strengthens proposals but is not required (unaffiliated applications may take longer to process)
• Clear demonstration of how the project meets all three criteria: importance, neglectedness, and tractability
• Interdisciplinary approaches welcomed—applications from outside computer science are encouraged

Common Challenges

• Out-of-scope proposals: Many 2024 applicants submitted proposals that didn't align with CAIF's specific mission
• Insufficient high-quality proposals: The Foundation had capacity to fund more projects but didn't receive enough strong applications
• Unclear fit: Applicants struggled to assess whether their work aligned with CAIF priorities under the previous broader scope

Strategic Advice

1. Carefully match your proposal to one of the eight specific research areas—this is the most important factor for success
2. Emphasize the cooperative aspects of your work and how it differs from dual-use capabilities
3. Be explicit about neglectedness—explain why this work won't happen without CAIF support
4. Think long-term—frame your work in terms of 10-20 year impact on advanced AI systems
5. For early-career researchers: Consider the early-career track, which assesses both project merit and career development potential
6. Prepare for iteration: The new process includes potential for interactive feedback, so be responsive and flexible

Key Takeaways for Grant Writers

• Scope is now precisely defined: Unlike 2024, the 2025 program requires applicants to select from eight specific research areas. Ensure perfect alignment before applying.

• Two-stage process reduces burden: The pre-proposal stage means you won't invest significant time unless invited to submit a full proposal. Make your pre-proposal compelling and clearly aligned.

• The Foundation has funding capacity: With only 50% of 2024 funds distributed, CAIF is actively seeking strong proposals. The challenge is fit, not overwhelming competition.

• Emphasize differential progress: Show how your work advances cooperative capabilities specifically, not just general AI capabilities that could be dual-use.

• International and interdisciplinary welcome: Location and discipline are not barriers. Computer science background is helpful but not required.

• Early-career track is significant: Up to £100,000 for a single researcher is substantial funding for postdoctoral work, with emphasis on career development.

• Reduced feedback for rejections: For 2025, most rejected applicants will receive minimal feedback. Only strong proposals in priority areas or promising early-career researchers will receive detailed feedback, so ensure initial alignment is clear.

References

• Cooperative AI Foundation website: https://www.cooperativeai.com/foundation
• 2025 Research Grants page: https://www.cooperativeai.com/grants/2025
• “Updates to the CAIF Grant Program in 2025”: https://www.cooperativeai.com/post/updates-to-the-caif-grant-program-in-2025
• “Grants Awarded by the Cooperative AI Foundation”: https://www.cooperativeai.com/post/grant-summaries
• UK Charity Commission Register: https://register-of-charities.charitycommission.gov.uk/charity-search/-/charity-details/5185411
• Allan Dafoe's statement on AI governance: https://www.allandafoe.com/opportunity
• CAIF application portal: https://cooperativeai.smapply.io/