Anthropic and OpenAI are privately fighting to control what external safety evaluators can publish. While both companies publicly promised independent reviews, they fear outside inspectors could leak trade secrets to commercial rivals.
FireBrief
FireBrief