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

Görsel: TechCrunch