Over 100 researchers and leaders signed an open letter. It calls on all frontier AI companies to embed independent evaluators. Evaluators must be independent, transparent, and shielded from retaliation. They must have access equal to senior employees. The letter lists five minimum conditions for credible evaluations. It cites the AEF-1 standard as an early example. Embedded evaluations should complement broader external oversight. The letter was published on September 18, 2026.
What it unlocks
Frontier AI companies can now embed independent evaluators under standardized conditions.
- 100+ signatories
Sources