People
Updated 05/18/26Advisor
Program Associate
Program Associate
Program Associate
Program Associate
Program Associate
Funding Details
Updated 05/18/26- Annual Budget
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
Org Details
Updated 05/18/26The AI Risk Mitigation Fund (ARM Fund) is a nonprofit grantmaking organization dedicated to reducing catastrophic risks from advanced artificial intelligence. Launched in December 2023, the ARM Fund was spun out of the Long-Term Future Fund (LTFF), building on the LTFF's track record of hundreds of grants totaling over $20 million in AI risk mitigation over five years. The ARM Fund operates across three primary focus areas. First, it funds technical AI alignment research, supporting work on mechanistic interpretability to develop deeper understanding of AI systems and scalable oversight methods enabling human supervision of advanced AI capabilities. Second, it supports AI policy and governance initiatives designed to ensure that government regulation and corporate standards appropriately guard against catastrophic risks. Third, it invests in building AI safety research capacity, recognizing that while tens of thousands of researchers work on AI capabilities, fewer than a thousand focus on AI safety. The fund's team of five Program Associates includes Lawrence Chan (researcher at ARC Evals), Oliver Habryka (co-founder of Lightcone Infrastructure and LessWrong), Lauro Langosco (PhD student at University of Cambridge), Thomas Larsen (Executive Director of the Center for AI Policy), and Caleb Parikh (who leads Effective Altruism Funds and manages the Long-Term Future Fund). They are supported by three Fund Advisors: Aviv Ovadya (Research Fellow at newDemocracy and founder of the AI and Democracy Foundation), Sam Bowman (Member of Technical Staff at Anthropic and NYU professor), and Adam Gleave (CEO of FAR AI). The ARM Fund is legally a project of Effective Ventures Foundation, a registered 501(c)(3) in the US and charity in the UK. One hundred percent of donations (minus payment processor fees) go directly to grantees, as the fund raises overhead and staff costs separately. The fund takes a hits-based approach to grantmaking, similar to venture capital, where a project with a small chance of success but massive potential impact may still receive funding. Each grant application is scored by fund managers on a scale from -5 to +5, with grants approved when average scores exceed a threshold typically set between 2.5 and 3.
Theory of Change
Updated 05/18/26The ARM Fund believes transformative AI is likely to be developed in the coming years or decades, and that current efforts to manage its risks are severely inadequate relative to the scale of potential catastrophe. Their theory of change operates on three fronts: funding technical alignment research to make AI systems more interpretable, steerable, and subject to human oversight before dangerous capabilities emerge; supporting policy work to establish government regulations and corporate standards that guard against catastrophic risks; and expanding the pipeline of AI safety researchers to close the talent gap between capability development and safety work. By taking a hits-based grantmaking approach that funds high-risk, high-reward projects, they aim to identify and support the most impactful interventions in AI safety, even when individual grants may have uncertain outcomes.
Grants Received
Updated 05/18/26Projects– no linked projects
Updated 05/18/26Discussion
Key risk: Because the ARM Fund’s people, networks, and grantmaking approach substantially overlap with LTFF/SFF and the EA safety ecosystem, the marginal counterfactual impact may be low and selection may inherit the same blindspots.
Case for funding: They offer high-leverage, hits-based regranting spun out of LTFF, with experienced evaluators across technical and policy (e.g., ARC Evals, CAIP, FAR AI, Anthropic) who can quickly deploy flexible low–six-figure grants to overlooked, high-upside AI safety projects.