FrontierSeptember 18, 2026via CNBC TechnologyCONFIRMED · 2 SOURCES · CROSS-PILLAR: FRONTIER+WORK

Anthropic and OpenAI need truly independent safety evaluators, experts say in public letter

Why it matters

Safety evaluation credibility is becoming a competitive and regulatory flashpoint. Labs self-evaluating their models creates conflicts of interest that practitioners and regulators are now refusing to accept—this could reshape how frontier capability claims get validated.

Key signals

  • Coalition of 100+ AI experts signed public letter
  • Target: Anthropic, OpenAI, and other foundation model labs
  • Core demand: independent and transparent safety evaluations
  • Published Sept 18, 2026 on CNBC
  • Implication: self-evaluation by labs is no longer credible to expert consensus
  • Targets Anthropic, OpenAI, and other foundation model labs
  • Core demand: independent safety evaluators, not lab-conducted evals
  • Published September 18, 2026
  • Governance/transparency issue tied to frontier model releases

The hook

100+ AI experts demand independence: foundation labs can't grade their own safety work anymore.

A coalition of over 100 AI experts are urging independence and transparency from Anthropic, OpenAI and other foundation model labs to conduct evaluations.

The week's key stories, every Friday.

ONE BRIEFING · EVERY FRIDAY · FREE

Free. Unsubscribe anytime.

Anthropic and OpenAI need truly independent safety evaluators, experts say in public letter | KeyNews.AI