The Open-Source AI Stack
RSS

Grants · Program · Global

SFF-2025 AI safety allocations

Approximately $29M of SFF's $34.9M 2025 round directed to AI safety and governance work, with smaller allocations to biosecurity, longevity, forecasting, math, and EA infrastructure.

The Survival and Flourishing Fund's 2025 grant round recommended approximately $34.33 million total, with approximately $29 million directed to AI safety and governance work, approximately $2.5 million to biosecurity, and the balance across fertility, longevity, forecasting, mathematics research, memetics, EA community building, and non-AI/bio global catastrophic risk work. The 2025 round materially exceeded SFF's pre-round $10M to $20M estimate. Funder Jaan Tallinn was the principal funder; SFF previously distributed funds from Tallinn alongside Jed McCaleb in prior rounds.

SFF operates through what it calls the S-Process, short for Simulation Process. Twelve recommenders across three tracks (a Main Track, a Fairness Track, and a Freedom Track) work through a structured procedure in which each recommender constructs a marginal-utility function over potential grantees and then a simulation step combines these utility functions into a final allocation. The mechanism is meant to extract a recommended grant vector from heterogeneous evaluator preferences without a single chair making the final call. The S-Process replaces the conventional program-officer model used by most foundations.

Named SFF-2025 grantees include MIRI (with the first $1.6M of its 2025 fundraiser matched 1:1 via the SFF grant), Center for AI Safety (approximately $289K to CAIS plus $772K to the action fund), and METR (formerly ARC Evals), per EA Forum and 80,000 Hours analyses of the round. The full recommendations list runs to dozens of organizations covering technical alignment research (MIRI, FAR.AI-adjacent groups), evaluation infrastructure (METR), governance (CAIS), and field-building.

Within the stack, SFF-2025 is a major source of unrestricted operating support for the AI safety and governance organizations at the safety-guardrails, governance, and evaluation meta-layers. The S-Process structure makes it more idiosyncratic than program-officer-led foundations, with the practical effect that newer or smaller groups can be funded without a centralized program portfolio.

Recipient

Multiple AI safety and governance organizations

Funder

Survival and Flourishing Fund · foundation · Global

Distributes funds from Jaan Tallinn and other donors to projects working on long-term survival and flourishing of sentient life, with heavy weight to technical AI safety and AI governance.

Primary source

https://survivalandflourishing.fund/2025/recommendations

Additional sources