Musk backs cross-testing AI rivals' safety harnesses, says Anthropic cares more than OpenAI
elonmusk · x · 2026-09-15
Elon Musk endorsed a mutual-testing scheme: every AI lab has a test harness probing models for bioweapon/nuclear risk and deliberate deception, and labs should apply each other's harnesses via pre-release API access — the only approach he thinks all parties, including China, could agree to. He also said Anthropic takes AI safety more seriously than OpenAI.
Related event: Musk Proposes AI Labs Cross-Audit Each Other's Safety Tests(2 posts)→
More from Companies & People
- Rishi Bommasani: The real cost of joining a frontier lab is losing your public voice — RishiBommasani · 2026-09-15
- NVIDIA Seattle DGX Spark hack winners: offline survival app LifeKit and wearable assistant Memo — NVIDIA Developer · 2026-09-15
- Palantir and NVIDIA launch Sovereign AI OS Reference Architecture for enterprises — arieljalali · 2026-09-15
- Blogger claims METR is on Anthropic's payroll via $7.7B stock pile — "AI doom is a business model" — ayushtweetshere · 2026-09-15
- Musk & Shotwell on All-In: AI risks, Starship, space data centers and a possible SpaceX-Tesla merger — elonmusk · 2026-09-15
- AI bootcamp claim of 21 activities doesn't match the 18 listed on its site — benjaminjriley · 2026-09-15