Research institute · Research orgs · Est. 2022
Forecasting Research Institute (FRI) is a Delaware-incorporated 501(c)(3) nonprofit research organization that develops and applies forecasting methods to improve decision-making on high-stakes issues. Its work builds on the forecasting accuracy standards associated with Philip Tetlock’s earlier “first generation” forecasting research, and FRI extends these foundations by running large-scale forecasting studies focused on topics including AI, biosecurity, cybersecurity, and geopolitics, then translating results for policymakers, researchers, and the public.
Strategically, FRI sits at the intersection of forecasting methodology, calibration and incentive design, and applied risk analysis. A major technical differentiator is its focus on “resolution-ready” forecasting—designing questions and study structures that can be scored as events unfold—and on generating comparative benchmarks for AI forecasting (e.g., ForecastBench) alongside expert forecasting panels (e.g., LEAP). FRI also emphasizes transparency about governance and funding support. Its public transparency page lists grantmaking donors who have contributed $10,000+ since October 2022 and describes selected collaborations whose details are restricted until research is public.
A study to better understand the effectiveness of Anthropic's AI Safety Level 3 measures at reducing biosecurity risks.
Forecasting AI Cyber Risks and Capabilities: Results of a 2025 Pilot StudyA pilot study investigating how AI capabilities may affect near-term cybersecurity risk, focusing on two high-impact cyberattack pathways: data-damaging worm attacks and cyberattacks against the U.S. electrical grid.
Measuring Judgment Quality in Natural-Language Explanations: Evidence from Forecasting TournamentsWe introduce EQMs, a scalable, interpretable method for extracting judgment-relevant information from written rationales.
Near Term Xpt AccuracyExistential Risk Persuasion TournamentLongitudinal Expert Ai Panel LeapAi Needs Fewer Prophets And More PredictionsPsychometric Properties Of Probability And Quantile ForecastsProject Improbable Improving Low Probability JudgmentsReciprocal Scoring Forecasting Unanswerable QuestionsImproving Judgments Of Existential RiskLongrange Subjective Probability Forecasts Of Slowmotion Variables In World PoliticsFRI posts a forecasting study in which experts and superforecasters predict semiconductor/software stock indices, data-center investment, and revenue growth at OpenAI and Anthropic, framed as a “big if” regarding whether recent growth curves continue.
Forecasting the Impacts of Anthropic's ASL-3 Safeguards on Biosecurity RisksFRI posts results from a forecasting study about expert views on the effectiveness of Anthropic’s AI Safety Level 3 (ASL-3) Deployment Standards at reducing biosecurity risk, funded by Anthropic PBC, including survey details and scenario framing.
Experts Forecast Rapid AI Progress Could Bring Health and Wealth Without HappinessFRI posts highlights from LEAP Wave 10 focusing on potential benefits of AI (health, longevity, wealth) and explores experts’ forecasts on how such gains might affect life satisfaction.
Forecasting AI Cyber Risks and CapabilitiesFRI posts results from a 2025 pilot study (focused on 2026) surveying superforecasters and cybersecurity experts on how AI capabilities may affect near-term cybersecurity risk along selected high-impact cyberattack pathways.
AI models have likely reached parity with superforecasters on ForecastBenchFRI posts new ForecastBench leaderboard results (including Cassi AI, xAI, and Google DeepMind submissions) and interprets statistical comparisons as suggesting parity between some AI systems and superforecasters on forecasting accuracy.
FRI posts an analysis of whether performance on easy-to-resolve calibration/proxy questions correlates with forecasting accuracy on a harder-to-evaluate target (a wet-lab molecular biology RCT forecast).