Anthropic and OpenAI pledge to embed third-party safety evaluators, but true independence remains an open question
RebeccaBellan · x · 2026-09-17
In a lengthy essay, Anthropic CEO Dario Amodei proposed embedding third-party evaluators inside all frontier AI companies, empowered to report safety incidents, assess model alignment, and publish unvarnished findings. Anthropic committed to giving evaluators like METR and Redwood Research unprecedented access, and Sam Altman said OpenAI would do the same — a potentially profound shift in how the industry works with outside research groups.
Evaluators who spoke to TechCrunch broadly welcomed the proposal but say key details must be worked out — and ideally backed by legislation — before they can know whether they'll act as truly independent watchdogs or as vendors operating on the AI companies' terms.
More from Companies & People
- Sam Altman: AI query energy cost fell 1,000x-100,000x in 3-4 years — rohanpaul_ai · 2026-11-02
- Enterprise AI's real game: getting companies to spend far more, not less — inductionheads · 2026-11-02
- Sam Altman meets Samsung executives in Korea to deepen memory chip and AI infrastructure cooperation — pstAsiatech · 2026-11-02
- Tailscale called a 'single point of failure' for startups as acquisition chatter splits devs — TheZachMueller · 2026-11-02
- Ex-Shopify engineer: the smartest people I knew all left for frontier AI labs — smlpth · 2026-11-02
- Harvard releases 6 free AI lectures covering GenAI, prompt engineering and RAG — Aiden_Tech_Ai · 2026-11-02