Mandatory 4-month expert risk evals: Anthropic says yes, OpenAI says no
Hesamation · x · 2026-09-03
On the proposal that "frontier AI models must be evaluated every 4 months by independent experts for catastrophic risks," the reactions diverge sharply: OpenAI flatly rejects it, Google calls it excessive, while Anthropic is eager to sign on. The contrast underscores Anthropic's ongoing effort to differentiate itself on safety.
More from Safety
- NVIDIA open-sources a tool that scans AI agent skills for security risks before you run them — Roger_M_Taylor · 2026-09-03
- 67% of 2025 automotive cyber incidents involve back-end servers, says Upstream — moniquejmorrow · 2026-09-03
- "Robots should be air-gapped systems": a security argument for embodied AI — mallow610 · 2026-09-03
- Sovereign AI's water problem: a 1,024-GPU UAE cluster could drink 30M+ liters a year — Arc_bong · 2026-09-03
- More AI-Generated Fake Journalists Unearthed for Press Gazette — RobWaugh74 · 2026-09-03
- Guardian podcast: chatbots, sycophancy and the AI mental-health rabbit hole — nordicinst · 2026-09-03