Survival and Flourishing Fund (SFF)

Quick Assessment

Organization Details

Overview

The Survival and Flourishing Fund (SFF) is the second-largest funder of AI safety research after Coefficient GivingOrganizationCoefficient GivingCoefficient Giving (formerly Open Philanthropy) has directed $4B+ in grants since 2014, including $336M to AI safety (~60% of external funding). The organization spent ~$50M on AI safety in 2024, w...Quality: 55/100, having distributed over $100 million since beginning grantmaking in 2019. Financed primarily by Jaan Tallinn, the Skype and Kazaa co-founder with an estimated net worth of approximately $900 million, SFF uses a distinctive algorithmic mechanism called the "S-process" to allocate grants based on recommendations from multiple advisors.

SFF originated from the Berkeley Existential Risk Initiative (BERI) in 2019 as a way to continue BERI's grantmaking activities while allowing BERI to focus on its core mission of operational support. Initially funded with approximately $2 million from BERI (itself funded by Tallinn), SFF has grown dramatically: from $2 million distributed in 2019 to $34.33 million in 2025.

SFF's focus has increasingly centered on AI safety as the field has grown. In 2025, approximately 86% of grants went to AI-related projects, up from roughly 50% in 2019. This reflects both Tallinn's longstanding concern about AI existential risk and the growing urgency perceived in the field. The fund supports a diverse portfolio ranging from technical research organizations (MIRIOrganizationMachine Intelligence Research InstituteComprehensive organizational history documenting MIRI's trajectory from pioneering AI safety research (2000-2020) to policy advocacy after acknowledging research failure, with detailed financial da...Quality: 50/100, METROrganizationMETRMETR conducts pre-deployment dangerous capability evaluations for frontier AI labs (OpenAI, Anthropic, Google DeepMind), testing autonomous replication, cybersecurity, CBRN, and manipulation capabi...Quality: 66/100, FAR AIOrganizationFAR AIFAR AI is an AI safety research nonprofit founded in July 2022 by Adam Gleave (CEO) and Karl Berzins (Co-founder & President). Based in Berkeley, California, the organization conducts technical res...Quality: 76/100) to policy groups (Center for AI Policy, GovAIOrganizationGovAIGovAI is an AI policy research organization with ~40-45 staff, funded primarily by Coefficient Giving ($1.8M+ in 2023-2024), that has trained 100+ governance researchers through fellowships and cur...Quality: 43/100) and field-building initiatives (SERI MATS, 80,000 HoursOrganization80,000 Hours80,000 Hours is the largest EA career organization, reaching 10M+ readers and reporting 3,000+ significant career plan changes, with 80% of $10M+ funding from Coefficient Giving. Since 2016 they've...Quality: 45/100).

The S-process mechanism distinguishes SFF from traditional foundations. Rather than having a single decision-maker or voting committee, SFF uses multiple "recommenders" (typically 6-12 per round) who express their funding preferences as mathematical utility functions. An algorithm then computes final allocations that respect funders' meta-preferences about which recommenders to trust on which topics. Critically, the system is designed to favor funding projects that at least one recommender is excited about, rather than projects that achieve consensus approval.

2025 Grant Round

SFF's 2025 grant round distributed $34.33 million across dozens of organizations, significantly exceeding the initial $10-20 million estimate. The round featured three specialized tracks: the Main Track (6 recommenders, $6-12M), the Freedom Track (3 recommenders, $2-4M), and the Fairness Track (3 recommenders, $2-4M). In total, twelve recommenders participated in evaluating applications for funder Jaan TallinnPersonJaan TallinnProfile of Jaan Tallinn documenting $150M+ lifetime AI safety giving (86% of $51M in 2024), primarily through SFF ($34.33M distributed in a 2025 grant round). Co-founded CSER (2012) and FLI (2014),...Quality: 53/100.

2025 Breakdown by Cause Area

Notable 2025 AI Safety Grantees

2025 Matching Pledge Program

New to 2025, SFF introduced a Matching Pledge Program designed to diversify the funding landscape and increase grantee independence. Matching Pledges are commitments by funders to match outside donations at specified rates (0.5x, 1x, 2x, or 3x) up to pledged amounts. Organizations that opted into the program received algorithmic boosts to their evaluations, factoring in the expected leverage from external donors.

The goals of the Matching Pledge Program include:

• Diversifying funding sources beyond SFF
• Encouraging other donors to give more
• Increasing fundraising robustness and independence of grantees
• Reducing single-funder dependency risk

Non-AI Existential Risk (~14% / ≈$5M)

The S-Process Mechanism

The S-process ("S" stands for "Simulation") is SFF's distinctive grant allocation mechanism, co-developed by Andrew Critch, Jaan TallinnPersonJaan TallinnProfile of Jaan Tallinn documenting $150M+ lifetime AI safety giving (86% of $51M in 2024), primarily through SFF ($34.33M distributed in a 2025 grant round). Co-founded CSER (2012) and FLI (2014),...Quality: 53/100, Oliver Habryka, Kevin Arlin, and Jason Moggridge. Unlike traditional grantmaking where a committee votes or a single program officer decides, the S-process uses mathematical preference functions and an optimization algorithm to allocate funding.

flowchart TD
  subgraph Inputs["Inputs"]
      APPS[Applications<br/>Organizations submit funding requests]
      RECS[Recommenders<br/>4-12 experts per round]
      DONORS[Funder Meta-Preferences<br/>Trust weights per recommender]
  end

  subgraph Process["S-Process Simulation"]
      EVAL[Marginal Value Functions<br/>Each recommender specifies utility<br/>curves for each application]
      DISC[Discussion Rounds<br/>Multiple meetings to refine<br/>evaluations and share information]
      ALGO[Allocation Algorithm<br/>Cycles through recommenders,<br/>each allocates next $1K to<br/>highest-value application]
  end

  subgraph Output["Output"]
      GRANTS[Final Grant Amounts<br/>Published with full transparency]
      MATCH[Matching Pledges<br/>Optional leverage for grantees]
  end

  APPS --> EVAL
  RECS --> EVAL
  DONORS --> ALGO
  EVAL --> DISC
  DISC --> ALGO
  ALGO --> GRANTS
  ALGO --> MATCH

  style Inputs fill:#e6f3ff
  style Process fill:#ccffcc
  style Output fill:#ffffcc

How It Works: Step by Step

The S-process operates through a structured series of meetings and algorithmic simulations:

1. Application Submission: Organizations submit applications via the SFF Funding Rolling Application, describing their work, funding needs, and theory of change. Applications are accepted on a rolling basis throughout the year.

2. Recommender Selection: For each grant round, funders agree on a set of 4-12 "Recommenders" with relevant expertise. The 2025 round featured 12 recommenders across three tracks (Main, Freedom, Fairness), with 6 recommenders in the Main Track and 3 each in the specialized tracks.

3. Initial Evaluation: Recommenders review applications and specify marginal value functions for funding each organization. These functions express how much value the recommender places on each additional dollar granted to each applicant.

4. Discussion Meetings: Over a series of 4+ hour-long meetings, recommenders discuss applications, share information, and adjust their evaluations. According to recommender Zvi Mowshowitz, this typically involves "several additional discussions with other recommenders individually, many hours spent reading applications, doing research and thinking about what recommendations to make."

5. Funder Meta-Preferences: Funders (primarily Jaan Tallinn) specify their own value functions for deferring to each recommender. This creates a weighted influence system where funders can express differential trust in recommenders for different cause areas.

6. Algorithm Computes Allocations: The S-process algorithm runs a simulation that cycles through recommenders. In each cycle, each recommender allocates their next $1,000 to whichever application has the highest marginal value according to their function, given what's already been allocated. This continues until budgets are exhausted.

7. Final Adjustments: Funders review algorithmic recommendations and may make adjustments. They retain final authority over all grants and can make grants the algorithm didn't endorse based on information learned during the process.

8. Publication: Final grant amounts are published on the SFF website with full transparency about recipients and amounts.

Key Design Principle: Champion-Based Funding

The S-process is explicitly designed to favor funding things that at least one recommender is excited about, rather than things that every recommender is excited about. As SFF explains:

"The grant recommendations do not especially represent the 'average' opinion of the group in any sense."

This means organizations benefit most from having one or two strong champions among the recommenders, rather than achieving lukewarm consensus support. The cycling allocation mechanism ensures every recommender's top priorities get funded, with marginal decisions depending on finding enthusiastic backers.

Advantages of the S-Process

Criticisms and Limitations

Zvi Mowshowitz, who served as an SFF recommender, has written extensively about the process's limitations:

Jaan Tallinn: Primary Funder

Jaan Tallinn (born February 14, 1972) is an Estonian programmer, entrepreneur, and one of the most significant individual funders of AI safety research globally. His estimated net worth of approximately $900 million derives primarily from his founding role in two transformative tech companies: Kazaa (peer-to-peer file sharing) and Skype (sold to eBay in 2005, later to Microsoft for $8.5 billion in 2011).

Background and Career

Tallinn became concerned about AI existential risk after reading works by Nick BostromPersonNick BostromComprehensive biographical reference page on Nick Bostrom covering his major contributions (orthogonality thesis, instrumental convergence, Superintelligence), institutional history (FHI founding a...Quality: 25/100 and Eliezer YudkowskyPersonEliezer YudkowskyComprehensive biographical profile of Eliezer Yudkowsky covering his foundational contributions to AI safety (CEV, early problem formulation, agent foundations) and notably pessimistic views on AI ...Quality: 35/100. He describes himself as having "yet to meet anyone working at AI labs who thinks the risk of training the next-generation model 'blowing up the planet' is less than 1%." He was among the signatories of both the Future of Life Institute's 2023 open letter calling for a pause on training AI systems more powerful than GPT-4, and the Center for AI Safety's 2023 statement on mitigating extinction risk from AI.

Tallinn's AI Safety Investments and Philanthropy

Beyond grantmaking through SFF, Tallinn has made significant direct investments in AI safety:

2024 Philanthropy Overview

According to Tallinn's 2024 philanthropy overview, he allocated approximately $20 million through his personal foundation in 2024, focusing on long-term alignment research and field-building initiatives. This made him one of the largest individual AI safety donors that year. Key 2024 initiatives included funding the AI Futures Project / AI 2027 initiative.

In the broader context of AI safety funding, Tallinn's contributions through SFF and direct giving represent approximately 15-20% of total philanthropic AI safety funding, second only to Coefficient Giving. Analysis of the AI safety funding landscape estimates global AI safety research funding reached $110-130 million in 2024, with Tallinn contributing approximately $20 million through his personal foundation plus additional amounts through SFF.

Historical Grant Patterns

Grant Round Totals by Year

SFF's grantmaking has grown dramatically since its founding:

Note: 2023-H2 total includes Lightspeed Grants amounts that Jaan Tallinn requested be incorporated into the SFF announcement.

Evolution of Focus

Notable Cumulative Grantees

Organizations that have received significant SFF funding across multiple rounds:

SFF Timeline

Funding Ecosystem

flowchart TD
  subgraph Sources["Funding Sources"]
      JT[Jaan Tallinn<br/>Primary Funder<br/>≈$100M net worth]
      JM[Jed McCaleb]
      DM[David Marble<br/>Casey Family Foundation]
      SFC[Survival and Flourishing Corp]
  end

  subgraph Mechanisms["Funding Mechanisms"]
      SFF[SFF S-Process<br/>$14M in 2025<br/>Champion-based allocation]
      SPEC[Speculation Grants<br/>≈$16M budget<br/>1-2 week decisions]
      LIGHT[Lightspeed Grants<br/>≈$1M/round<br/>Days-weeks]
      MATCH[Matching Pledges<br/>2025+ program<br/>Leverage external donors]
  end

  subgraph Recipients["Major Recipient Categories"]
      AI[AI Safety Orgs<br/>86% of funding<br/>MIRI, METR, CAIS, etc.]
      BIO[Biosecurity<br/>7% of funding<br/>SecureBio, JHU CHS]
      OTHER[Other X-Risk<br/>7% of funding<br/>Nuclear, civilizational]
  end

  JT --> SFF
  JT --> SPEC
  JT --> LIGHT
  JM --> SFF
  DM --> SFF
  SFC --> SFF

  SFF --> AI
  SFF --> BIO
  SFF --> OTHER
  SPEC --> AI
  LIGHT --> AI
  MATCH --> AI

  style Sources fill:#e6f3ff
  style Mechanisms fill:#ccffcc
  style Recipients fill:#ffffcc

Speculation Grants Program

In addition to the main S-process rounds, SFF operates a Speculation Grants program for expedited funding. This addresses a key limitation of the S-process: its 3-6 month timeline can be too slow for time-sensitive opportunities.

How Speculation Grants Work

Key Features

Eligibility Gateway: Receiving a Speculation Grant of $10K+ guarantees eligibility for the next S-process round. This provides an entry path for organizations unknown to recommenders.

Speed vs. Information Trade-off: As the program notes, "to get money faster, you have to provide more information, not less." Applicants must submit full applications even for expedited funding.

S-Process Integration: If an organization receives a Speculation Grant and later receives an S-process recommendation, they only receive additional funds to the extent the S-process amount exceeds the Speculation Grant (avoiding double-counting).

2024 Example

In the 2024 round, $0.85M in funding was distributed previously through Speculation Grants, integrated into the total $19.86M round announcement.

Comparison with Other Funders

SFF operates alongside but independently from other major longtermist funders, each with distinct approaches and comparative advantages:

Funding Landscape Overview

Source: EA Forum analysis of AI safety funding

Comparative Characteristics

Strategic Position

SFF's Niche: SFF is often willing to fund organizations that other funders consider higher-risk or more speculative, making it an important source of support for early-stage research groups. The S-process's champion-based design means an organization can receive funding if even one recommender is strongly enthusiastic, whereas consensus-based approaches might reject the same application.

Post-FTX Importance: After the collapse of FTX and the Future Fund in late 2022, SFF became even more critical to the longtermist funding ecosystem. The Future Fund had been positioned as a major new funder with similar cause priorities; its disappearance increased reliance on SFF and Coefficient Giving.

LTFF Relationship: LTFF has received funding from both SFF and Coefficient Giving, making it partially downstream of these larger funders. About 40% of LTFF's 2022 funding came from Coefficient (then Open Philanthropy). LTFF typically makes smaller grants ($10K-$500K) compared to SFF's median ≈$100K and often funds individuals or very early-stage projects.

Lightspeed Grants: Also primarily funded by Jaan Tallinn, Lightspeed Grants focuses on even faster turnaround than SFF's Speculation Grants. The 2023-H2 round included $9.62M from Lightspeed Grants incorporated into the SFF announcement.

Application Process

How to Apply

Applications are submitted through the SFF Funding Rolling Application. A single submission requests consideration for both Speculation Grants (expedited) and the next S-process round.

Timeline

Eligibility Requirements

Tips for Applicants

Based on public information about successful grants and recommender commentary:

What Works:

1. Find a Champion: The S-process rewards having at least one recommender who is enthusiastic about your work. Being known to recommenders helps significantly.
2. Clear Theory of Change: Explain specifically how your work reduces existential risk, with logical chain from activities to impact.
3. Concrete Outputs: Describe specific deliverables and milestones rather than vague research directions.
4. Team Credibility: Highlight relevant experience, past work, and track record. Reference legible signals where possible.
5. Appropriate Ask Size: The process rewards asking for larger amounts, but ask for what you can actually absorb and deploy effectively.
6. Provide More Information: For faster funding (Speculation Grants), provide more detail, not less.

Potential Challenges:

• New organizations: Without existing relationships to recommenders, may need to go through Speculation Grants first
• Non-AI focus: With 86% of funding going to AI, non-AI projects face steeper competition
• Consensus-dependent projects: The champion-based model may disadvantage projects that are "good but not great" to everyone
• Limited feedback: Rejected applicants may not receive detailed explanations

2025 Tracks

The 2025 round featured three specialized tracks, and all eligible applications were evaluated in all tracks:

SFF explains the specialized tracks: "Fairness and freedom are values SFF considers crucial to humanity's survival and flourishing in the era of AI technology, especially now that leading experts in AI have acknowledged that AI presents an extinction-level threat to humanity."

Key Debates and Considerations

The Champion-Based Model

The S-process's design to fund projects with at least one enthusiastic recommender, rather than consensus picks, is both a strength and a debate point:

Arguments For:

• Surfaces innovative projects that might be filtered out by consensus processes
• Allows recommenders with specialized knowledge to back projects others don't understand
• Prevents "design by committee" homogenization of the funding portfolio
• Rewards organizations that build strong relationships with knowledgeable advocates

Arguments Against:

• May fund projects that are genuinely bad ideas one person happens to like
• Creates incentives to cultivate individual recommenders rather than build broad support
• Could lead to funding based on personal relationships rather than merit
• Makes the recommender selection process highly consequential

EA Ecosystem Influence

Zvi Mowshowitz has noted that despite no official relationship between SFF and Effective Altruism, "at least the SFF process and its funds were largely captured by the EA ecosystem. EA reputations, relationships and framings had a large influence on the decisions made." This raises questions about:

• Whether SFF provides genuine diversification from EA-aligned funders
• How organizations outside EA networks can access SFF funding
• Whether the 2025 Freedom and Fairness tracks genuinely broaden perspectives

Single-Funder Dependency

SFF is heavily dependent on Jaan Tallinn as its primary funder. While other funders (Jed McCaleb, David Marble) participate, Tallinn's ≈$900M net worth and commitment to AI safety are central to SFF's scale. This creates:

• Sustainability risk: SFF's future depends significantly on Tallinn's continued wealth and priorities
• Governance concentration: One person's views heavily shape funding direction
• Mitigation efforts: The 2025 Matching Pledge Program explicitly aims to diversify funding sources

AI Concentration Trade-offs

The shift from ~50% AI focus in 2019 to ~86% in 2025 reflects both genuine urgency and potential trade-offs:

• For: AI risk may genuinely be the most pressing x-risk; funding follows perceived importance
• Against: Biosecurity, nuclear risk, and other x-risks may be relatively underfunded; portfolio diversification has value under uncertainty

Strengths and Limitations

Organizational Strengths

Organizational Limitations

Sources and Citations

Primary Sources

• SFF Official Website
• S-Process Explanation
• 2025 Grant Round Application
• 2025 Grant Recommendations
• 2024 Grant Recommendations
• Speculation Grants Program
• SFF Announcements Archive

Recommender Commentary

• Zvi's Thoughts on the Survival and Flourishing Fund (SFF) - Zvi MowshowitzPersonZvi MowshowitzAI safety commentator and blogger. Author of 'Don't Worry About the Vase'. Former professional Magic: The Gathering player. Writes weekly AI updates. (LessWrong)
• Thoughts on the Survival and Flourishing Fund 2024 Round - Zvi Mowshowitz (Substack)
• SFF Speculation Grants as an Expedited Funding Source - EA Forum

Funding Analysis

• An Overview of the AI Safety Funding Situation - EA Forum
• SFF 2025 Funding by Cause Area - EA Forum
• Survival and Flourishing Fund Donations Made - Vipul NaikPersonVipul NaikVipul Naik is a mathematician and EA community member who has funded ~$255K in contract research (primarily to Sebastian Sanchez and Issa Rice) and created the Donations List Website tracking $72.8...Quality: 63/100

Jaan Tallinn

• Jaan Tallinn - Wikipedia
• Jaan Tallinn - CSER Profile
• Jaan Tallinn - Future of Life Institute
• Jaan Tallinn's 2024 Philanthropy Overview - LessWrong

Related Organizations

• Survival and Flourishing Corp - Manages S-process operations
• Berkeley Existential Risk Initiative (BERI) - SFF's origin organization

External Links

• SFF Website
• Grant Recommendations History
• S-process Explanation
• Speculation Grants
• Rolling Application Form
• Survival and Flourishing Corp
• EA Forum Topic: Survival and Flourishing