Survival and Flourishing Fund

The Survival and Flourishing Fund (SFF) is a US-based "virtual fund" that organises the collection and allocation of philanthropic donations to organisations concerned with the long-term survival and flourishing of sentient life. Launched in the third quarter of 2019, SFF emerged from a decision by the Berkeley Existential Risk Initiative (BERI) to hand off Jaan Tallinn's grantmaking to a separate vehicle so BERI could concentrate on its core mission of operational support for existential-risk researchers. Applications opened that autumn and grantmaking began in 2019 with approximately $2 million. Since then, SFF has organised approximately $152 million in philanthropic gifts and grants.

SFF holds a donor-advised fund (DAF) at the Silicon Valley Community Foundation through which most grants flow; a parallel entity, Survival and Flourishing Corp (SFC), processes Initiative Committee grants and some others directly. Ethan Ashkie serves as President and CEO of Survival and Flourishing Corp. The "virtual fund" framing signals that SFF is not an endowed foundation but a coordination infrastructure — a structured evaluation process and a DAF vehicle, rather than an institution with staff and endowment of its own.

Funders and scale

Jaan Tallinn — Estonian software engineer, co-founder of Skype and Kazaa — is the primary donor and guiding force behind SFF. A committed longtermist concerned about AI existential risk, Tallinn channels philanthropic resources through SFF at significant scale. Additional donors in recent rounds have included Jed McCaleb, Blake Borgeson, and David Marble via the Casey and Family Foundation, though Tallinn's contributions dominate the total.

From an initial $2 million in 2019, SFF's annual grantmaking has grown substantially: approximately $33 million in 2023, $41 million in 2024, and $34.33 million in 2025. Across its first six years it has organised approximately $152 million in total. AI safety now constitutes approximately 86% of the 2025 allocation, with biosecurity and other existential-risk categories accounting for the remainder — a concentration that gives SFF the profile of the second-largest dedicated AI safety funder globally, behind only Coefficient Giving.

The S-Process

SFF's primary grant-allocation mechanism is the S-Process — an algorithmic procedure co-developed by Andrew Critch, Jaan Tallinn, Oliver Habryka, Kevin Arlin, and Jason Moggridge. The S-Process runs once or twice a year: for each round, four to twelve volunteer recommenders review incoming applications and express their evaluations as marginal value functions — mathematical utility functions capturing how much value each recommender places on each additional dollar granted to a given applicant. Over multiple sessions, recommenders discuss applications and refine their functions. An allocation algorithm distributes funds by favouring proposals with at least one strong advocate over those with only diffuse lukewarm consensus, embedding a "champion-based" logic designed to surface high-conviction bets that might otherwise be averaged away. Funders retain final authority over whether to act on the recommendations.

The 2025 round introduced two specialist parallel tracks — a Fairness Track and a Freedom Track, each with three recommenders — running alongside the main six-recommender track, broadening the evaluation perspective across distinct philosophical priorities. SFF also operates Speculation Grants — smaller, faster grants made on a rolling basis and retroactively validated by subsequent S-Process rounds — and an Initiative Committee track in which Tallinn, a small group of advisors, and two to five anonymous voters make grants on their own initiative without requiring proposals.

Field-building and community grantees — the corpus connection

SFF's field-building line, operated within the AI safety portfolio, extends to grassroots AI-safety community hubs and capacity-building organisations — the layer of the AI-safety movement most directly within this corpus's remit. Singapore AI Safety Hub (SASH), the Singapore-anchored coworking space and ecosystem-builder whose mission is to develop a civil-society AI-safety practitioner community in Asia, received $300,000 plus a $36,000 matching pledge in the SFF-2025 round — one of the larger community-hub grants in the 2025 allocation. Grants in this line also include AI Safety Camp ($90,000 plus a $110,000 matching pledge in 2025) and Apart Research ($294,000 in 2024), two field-building programs that run distributed hackathons and research cohorts with a strong community-hub dimension.

Position in the corpus funder slice

Within this corpus's funder slice, SFF fills a slot adjacent to Coefficient Giving — the EA-aligned AI safety funder it most closely parallels — while differing from it in structure and scale. Coefficient Giving operates as a fully institutionalised foundation with a large professional staff, a formal cause-selection methodology applied across multiple cause areas, and over $650 million in 2024 grantmaking; SFF is a leaner virtual-fund coordination vehicle directed primarily by a single major donor and a rotating volunteer recommender pool, with no endowment of its own and roughly one-fifteenth the annual grantmaking scale. Where Coefficient Giving's AI safety programme runs through the Navigating Transformative AI fund with a formal portfolio logic, SFF's allocations emerge from a champion-based procedure that foregrounds individual recommender conviction over institutional-consensus gatekeeping.

Both funders support the grassroots AI-safety community-hub layer, but SFF's field-building grants reach international hubs that Coefficient Giving's portfolio is slower to cover directly: the Singapore AI Safety Hub grant in SFF-2025 represents a significant investment in the Asia-Pacific community-building layer. In that sense the two funders are complementary within the AI safety funding ecosystem — sharing the longtermist-EA philosophical frame but differentiated by governance model, scale, and the distinct selection logic their respective allocation mechanisms embed.