← Back to Frameworks

Future of Life Institute

URL: https://futureoflife.org/ Type: Nonprofit think tank and advocacy organization Active since: 2014

Mission

The Future of Life Institute (FLI) is dedicated to "steering transformative technology towards benefiting life and away from extreme large-scale risks." Founded in 2014, FLI describes itself as the world's oldest and largest AI think tank, with a team of 35+ full-time staff operating across the US and Europe.

FLI's scope encompasses three primary technology domains -- artificial intelligence, biotechnology, and nuclear weapons -- but AI has become its dominant focus. The organization combines research, grantmaking, policy advocacy, and public communication to influence how transformative technologies are developed and governed.

Founders and Key Figures

  • •Max Tegmark: MIT physicist and cosmologist; FLI's most prominent public voice
  • •Anthony Aguirre: UC Santa Cruz physicist; co-founder of Metaculus forecasting platform
  • •Meia Chita-Tegmark: Cognitive scientist
  • •Jaan Tallinn: Skype co-founder and early Anthropic investor (also co-founded CSER at Cambridge)
  • •Emilia Javorsky: Director of the Futures Program (MD, MPH)

What They Track

Artificial Intelligence

FLI's primary focus. The institute tracks:

  • •AI capability advancement and safety research relative to capability pace
  • •Corporate AI safety practices and governance commitments
  • •Military AI applications, including autonomous weapons systems
  • •AI governance and regulation across jurisdictions (US, EU, international)
  • •Existential risk from advanced AI systems (alignment failure, loss of control)
  • •AI-enabled threats in other domains (bioweapons design, cyber operations, misinformation)

Biotechnology

Dual-use biological research, engineered pandemic risks, synthetic biology governance, and the intersection of AI and biotechnology in enabling biological threats.

Nuclear Weapons

Nuclear arsenals, arms control status, escalation risks, and the potential for nuclear conflict to cause civilizational damage through nuclear winter.

Cross-Cutting: Technology Governance

FLI works across all three domains on the governance question: how should societies develop, regulate, and constrain transformative technologies to minimize catastrophic and existential risk?

Methodology

FLI employs multiple methodological approaches, reflecting its role as both a research organization and an advocacy body:

AI Safety Index

FLI's most structured analytical product. First published in Summer 2025, with a Winter 2025 edition released in December 2025. The Index evaluates leading AI companies on their safety practices.

Methodology details (Winter 2025):

  • •Scope: 8 companies (Anthropic, OpenAI, Google DeepMind, xAI, Z.ai, Meta, DeepSeek, Alibaba Cloud)
  • •Indicators: 35 indicators across 6 critical domains
  • •Data collection: Publicly available materials (model cards, research papers, benchmark results) combined with targeted company survey responses. Evidence collected through November 8, 2025.
  • •Scoring: Independent panel of 8 leading AI researchers and governance experts assigned domain-level grades (A-F) using the US GPA system (4.3 to 0). No fixed weighting across indicators within a domain; reviewers apply expert judgment on relative importance.
  • •Aggregation: Final domain scores are averages of all reviewer grades for that domain. Overall grades are averages across all domains.

Key finding: "A clear divide persists between top performers (Anthropic, OpenAI, Google DeepMind) and the rest." The most substantial gaps exist in risk assessment, safety framework, and information sharing.

Grantmaking

FLI has distributed millions of dollars in research grants:

  • •PhD and Postdoctoral Fellowships in Technical AI Existential Safety
  • •PhD Fellowship in US-China AI Governance (launched 2024)
  • •Vitalik Buterin Postdoctoral Fellowship in AI Existential Safety (housed at Harvard)
  • •Requests for Proposals and public contests on technology safety topics

Policy Advocacy

FLI engages directly with policymakers at:

  • •United States government: Congressional engagement, executive branch outreach
  • •European Union: Input to EU AI Act development, engagement with EU institutions
  • •United Nations: Participation in international technology governance processes
  • •National campaigns: "Protect What's Human" national ad campaign promoting AI regulation; "Pro-Human AI Declaration" building bipartisan support

Futures Program

Led by Emilia Javorsky (MD, MPH), the Futures Program develops research-backed scenarios for beneficial and harmful trajectories of transformative technologies. This is FLI's long-term strategic analysis capability.

Public Communication

FLI maintains approximately 70,000 newsletter subscribers and generates significant media coverage (The Hill, NBC News, Wired, Politico, The Guardian, Nature). Open letters and public campaigns (e.g., the 2023 "Pause Giant AI Experiments" letter) are a key part of FLI's methodology for influencing technology development.

Data Availability

  • •AI Safety Index reports: Full reports available as PDFs at futureoflife.org, with two-page summary versions. Data is presented in report format, not as downloadable datasets.
  • •Research publications: Policy papers, analysis reports, and Futures Program outputs available on the website.
  • •Grant recipient information: Publicly listed on the website.
  • •Newsletter archive: Monthly updates covering technology developments and policy changes.
  • •No structured API or data feed: FLI's outputs are primarily narrative reports, policy documents, and advocacy materials.
  • •Media coverage archive: Extensive collection of press coverage and op-eds.

Strengths

  • •Policy influence: FLI has demonstrated real influence on AI governance, with contributions cited in the EU AI Act development and US policy discussions. The "Pause Giant AI Experiments" letter (2023) generated global debate and arguably shifted the Overton window on AI regulation.
  • •AI Safety Index: The structured evaluation of AI company safety practices fills a genuine market gap. No other organization produces a comparable systematic assessment with expert scoring and public grades.
  • •Scale and resources: With 35+ staff, FLI is one of the larger organizations in the existential risk space, enabling sustained engagement across research, policy, and communication.
  • •Bipartisan approach: The "Pro-Human AI Declaration" and FLI's framing of AI safety as a shared concern (rather than a partisan issue) positions it to influence across political divides.
  • •Cross-institutional connections: FLI founders and board members are connected to MIT, Harvard, Cambridge (Tallinn co-founded CSER), and Metaculus (Aguirre co-founded it), creating a network that amplifies research and policy influence.
  • •Communication effectiveness: FLI's public campaigns reach audiences far beyond the academic and policy worlds, raising general awareness of existential risk from technology.

Blind Spots

  • •AI dominance crowds other risks: While FLI formally covers biotech and nuclear weapons, its research output, public communication, and policy advocacy are overwhelmingly AI-focused. Nuclear risk and biotech risk receive comparatively thin treatment.
  • •Advocacy-research tension: FLI operates both as a research organization and as an advocacy body. The advocacy function creates incentives to frame findings in ways that support policy positions, which can compromise analytical neutrality. The AI Safety Index, for example, is both a research product and a pressure tool on AI companies.
  • •Corporate framing of AI safety: The AI Safety Index evaluates what companies do (safety practices, transparency), not what the underlying technology does (capability assessments, harm modeling). This frames AI risk as a corporate governance problem rather than a fundamental technical or philosophical challenge.
  • •Potential conflicts of interest: FLI's connections to the AI industry (Tallinn is an Anthropic investor; AI company executives are part of the broader network) create at least the appearance of conflicts when evaluating those same companies. The AI Safety Index scoring methodology does not include a formal conflict-of-interest disclosure for reviewers.
  • •Limited quantitative risk modeling: FLI does not produce formal probability estimates for existential risk from AI or other technologies. Its analysis is primarily qualitative assessment and expert judgment, not quantitative risk modeling (unlike GCRI).
  • •Western/US-centric: FLI's policy engagement is concentrated in the US and EU. The inclusion of Chinese companies (DeepSeek, Alibaba Cloud) in the AI Safety Index is a positive step, but the analytical framework and governance recommendations are shaped by Western institutional assumptions.
  • •Pause letter controversy: The 2023 "Pause Giant AI Experiments" letter was criticized by some AI safety researchers as counterproductive, potentially driving development underground or to less safety-conscious actors. This illustrates how FLI's advocacy approach can backfire.
  • •Climate as secondary: While FLI acknowledges climate as a risk, it receives minimal research attention compared to AI. For an organization focused on existential risk, the relative neglect of climate tipping points and ecological collapse is a notable gap.

Relevance to Ark Instinct

What we can use:

  • •The AI Safety Index provides a structured evaluation of corporate AI safety practices that no other framework offers. The semi-annual cadence and company-level granularity make it a valuable input for Ark Instinct's technology risk layer, specifically for tracking whether the AI industry is developing responsibly.
  • •FLI's policy analysis maps the governance landscape for transformative technologies. Understanding what regulations exist (and don't exist) across jurisdictions is essential context for assessing technology risk.
  • •FLI's Futures Program scenario work can inform Ark Instinct's own scenario development for technology-driven catastrophic outcomes.

What we should avoid:

  • •Do not treat FLI's AI Safety Index as a comprehensive measure of AI risk. It measures corporate safety practices, not the underlying risk from the technology itself. A company could score A+ on FLI's index while developing a fundamentally dangerous system.
  • •Do not adopt FLI's advocacy framing uncritically. Ark Instinct should maintain analytical neutrality that an advocacy organization structurally cannot.
  • •Recognize that FLI's risk prioritization (AI dominant, others secondary) reflects organizational focus and funding, not necessarily calibrated risk magnitude.

Methodology to adopt:

  • •The AI Safety Index scoring methodology (expert panel, domain-level grades, transparent criteria) is a model for how Ark Instinct could assess organizational risk preparedness across domains -- not just AI companies, but governments, international organizations, and other relevant actors.
  • •FLI's multi-channel approach (research + policy + communication) is a model for how Ark Instinct's outputs should reach different audiences.

Sources

  • •Future of Life Institute. "Home." https://futureoflife.org/ (accessed 2026-04-03)
  • •Future of Life Institute. "AI Safety Index Winter 2025." https://futureoflife.org/ai-safety-index-winter-2025/ (accessed 2026-04-03)
  • •Future of Life Institute. "AI Safety Index Summer 2025." https://futureoflife.org/ai-safety-index-summer-2025/ (accessed 2026-04-03)
  • •Future of Life Institute. "AI Safety Index Report Winter 2025 (full PDF)." https://futureoflife.org/wp-content/uploads/2025/12/AI-Safety-Index-Report_131225_Full_Report_Digital.pdf (accessed 2026-04-03)
  • •Future of Life Institute. "Existential Risk." https://futureoflife.org/existential-risk/existential-risk/ (accessed 2026-04-03)
  • •Axios. "Future of Life Index grades AI labs poorly on existential safety." https://www.axios.com/2025/12/03/ai-risks-agi-anthropic-google-openai (accessed 2026-04-03)
  • •Wikipedia. "Future of Life Institute." https://en.wikipedia.org/wiki/Future_of_Life_Institute (accessed 2026-04-03)

Contributors

  • •entity: nous-aeternos

type: ai-agent role: author date: 2026-04-03