For frontier labs and AI safety researchers: a working route from technical findings to the doctrine, acquisition, and operational decisions that determine how American AI gets fielded — with editorial standards on both sides.
Independent analysis. Editorial standards on both sides. Attribution by agreement.
01 · Independent Evaluation
Rigorous, non-commercial assessment of frontier model capabilities in scenarios relevant to national security decision-makers. Findings produced for, and trusted by, both sides of the table — because the evaluator has no equity in the outcome.
Read more →02 · Translation
Turning research outputs into the forms operators, program offices, and acquisition leaders can act on, without losing technical fidelity. Editorial standards on both sides; attribution by agreement.
Read more →03 · Convening
A facilitated venue for direct conversation between frontier labs and the National Security Enterprise. Discretion on our partners' terms. Attribution by agreement.
Read more →Frontier Security Institute at the Center for AI Safety focuses on the National Security Enterprise. Our work is grounded in CAIS's mission of reducing societal-scale risks from artificial intelligence.
safe.aiEvaluations, red-teams, joint convenings, technical advisory — if your work shapes what frontier AI can do, ours determines how it gets used.