Organizations and teams whose primary mission is AI safety, alignment, or AI governance.
Includes safety teams inside large AI labs (estimated FTEs for safety-relevant staff only),
university research groups with dedicated AI safety funding, and government AI safety bodies.
Excludes broader responsible AI, AI ethics teams at tech companies, general ML fairness research,
and AI regulation bodies not focused on catastrophic/existential risk.
FTE estimates are from Stephen McAleese's AI safety organization dataset, as of 2025-09-27 (a few orgs refreshed individually since carry their own newer date).
Box area =
Color:
501(c)(3) 501(c)(4) otherGrouped by subfield. Boxes shown only where the metric is known.
Has budget data Missing budget data
Box area = cycle spend, to scale. Pro-industry money dwarfs pro-safety. Labels show actual $. Pro-industry Pro-safety
Box area = total disbursed. Grouped by funder type.
Scope caveat
"AI safety" here means organizations focused on alignment, safety research, and AI governance around catastrophic/existential risk. A 3x scope expansion (e.g. including all responsible-AI and ML fairness work) could increase headcount estimates by ~10x. These comparisons reflect the narrow scope defined above.