Anthropic and OpenAI embed independent safety reviewers in labs

Anthropic and OpenAI announced plans to station independent safety evaluators inside their facilities for unprecedented access to AI development, with researchers welcoming the move but warning transparency remains essential as of September 18, 2026. The initiative seeks to balance operational access with meaningful oversight.