Anthropic lets METR audit its AI safety systems independently

Anthropic gave six METR employees employee-level access to its models and safety tools. They can publish safety findings without needing Anthropic's approval first. Feels like a genuine step toward independent AI auditing.