ANTHROPIC EXPANDED INDEPENDENT ACCESS TO ITS AI SAFETY WORK

Summary from the TLDR AI newsletter

Anthropic began giving METR employee-like access to models, training information, safeguards, and other internal material so it could independently assess frontier-model risks and publish its conclusions.