Anthropic and OpenAI want to embed safety evaluators. Will they really be independent?

Anthropic and OpenAI plan to embed safety evaluators in their labs, but experts warn that meaningful oversight requires transparency, independence, and regulation.