Center for AI Safety About (web)
Citation
Center for AI Safety (CAIS). “About us.” Web page, date not stated. https://safe.ai/about. Read 2026-11-05.
Bottom line
- CAIS is “an AI safety non-profit” that does research, field-building and advocacy. Field-building includes “Providing top researchers with compute and technical infrastructure”.
- It says it has given compute to “>100 researchers working on safety-related research at leading universities since March 2023”. Its Cloud and DevOps team runs a cluster that “supports c. 20 research labs”.
Background
The page gives the mission, three lines of work with activity lists, impact figures, a four-step project method and the team structure. It names two leaders: Dan Hendrycks, Executive & Research Director, and Oliver Zhang, Managing Director. The compute program has its own page: Center for AI Safety Compute Cluster (web).
Key claims
- “CAIS reduces societal-scale risks from AI through research, field-building, and advocacy.” (What we do)
- “CAIS grows the AI safety research field through funding, research infrastructure, and educational resources.” Activities include “Providing top researchers with compute and technical infrastructure” and “Running multidisciplinary fellowships focused on AI safety”. (Field-building)
- “Many researchers outside large tech companies are unable to work on understanding potential risks from cutting-edge AI models due to the cost of the high-end computing hardware required. We have provided access to our computing infrastructure to >100 researchers working on safety-related research at leading universities since March 2023.” (CAIS Impact)
- “A further 140 papers have been produced by CAIS partners using CAIS’ compute cluster or participating in our ML Safety workshops” (CAIS Impact)
- “Cloud and DevOps: Oversees the compute cluster, which supports c. 20 research labs.” (Center for AI Safety Teams)
Evidence and method
The organisation’s own page. The figures are self-reported. The headline numbers do not always match the text under them: the heading says 500 students trained, the text says “almost 300”.
Limits
No budget, funder list or hardware detail on this page. The page does not say how researchers are selected for compute.