AI Safety Watchdog METR Gains Unprecedented Access to Anthropic's Frontier Models

The AI safety research organization METR has secured an unprecedented level of access to Anthropic, a leading developer of advanced AI models. This collaboration grants METR deep insights into Anthropic's systems, including employee-level access, enabling comprehensive independent assessments of their AI models. The focus of these assessments will be on identifying potential risks, analyzing past incidents, and evaluating Anthropic's safety protocols.

Crucially, Anthropic has agreed to waive its right to censor METR's findings, allowing the research group to publish its evaluations without restriction. This agreement marks a significant advancement in AI transparency and underscores a commitment to rigorous, independent safety verification processes within the rapidly evolving field of artificial intelligence.