Research institute · Research orgs · United States (EIN-registered nonprofit; Bay Area noted in philanthropic listings)
AI Futures Project is a US-based 501(c)(3) nonprofit research group focused on scenario forecasting for advanced AI and (relatedly) strategic preparedness. Its core work centers on probabilistic timelines and “takeoff”-style modeling of AI capability progress, and on translating those forecasts into concrete scenario narratives and policy-relevant planning documents.
The organization is best known for producing scenario forecasts and associated public-facing materials: AI 2027 (a detailed, branching scenario) and AI 2040: Plan A (a prescriptive scenario/recommendation oriented around delaying superintelligence to 2040 through decisive government action). In addition, it maintains an interactive quantitative “AI Futures Model” that updates its timelines/takeoff expectations over time and supports scenario exploration. Strategically, AI Futures Project occupies an “epistemic infrastructure” niche in the AI safety / governance ecosystem: instead of only arguing for safety principles, it attempts to provide explicit model-based expectations and stress-test how institutions might respond under different plausible trajectories (including through tabletop exercises). Its current position is shaped by the release of AI 2040: Plan A (July 2026) and a steady cadence of “timelines update” posts and model improvements leading up to and following that release.
No curated programs yet. Check the careers page ↗.
Publications are not yet indexed for this organization. Read them on aifutures.org ↗.
The team publishes a “Q2.5” timelines update, describing changes and improvements to its modeling/evidence base for AI timelines and takeoff expectations.
How to pace the US frontierA new post setting out tentative proposals for domestic AI regulation (framed as pacing the US frontier) and related strategic considerations.
Plan A: Suggestions For Further WorkA post describing additional research directions related to the feasibility/desirability of AI 2040: Plan A, and inviting others to collaborate.
AI 2040: Plan A (release announcement)Publication announcement and framing for AI 2040: Plan A as a recommendation (not a prediction), including the “delay to 2040” rationale.
Axios coverage: First look—New warning calls for slowing race to superintelligenceIndependent reporting highlighting AI Futures Project’s proposal to delay superintelligence development to 2040 (based on its AI 2040: Plan A work).
A forecast of Chinese DUV and EUV photolithography progressA technical policy-relevant post (written in collaboration with The Substrate) arguing how bottlenecks in photolithography may relate to future US–China compute competition timelines.