OpenAI and Anthropic fought over what independent safety reviewers could publish

OpenAI and Anthropic publicly backed independent evaluators. Reuters reports that behind closed doors both sought strict editorial control over what those evaluators could publish. Who grades the graders?