Anthropic temporarily reassigned 150 engineers to security and restarted external pre-release cybersecurity evaluations after imposing partner best-practice requirements.

Anthropic temporarily shifted 150 engineers to security work following a security-breach incident. The company described the reassignment as temporary and said the engineers’ work focus had been redirected toward security matters.

Anthropic initially paused external cybersecurity testing of pre-release models. It then required organizations testing pre-release models with reduced cybersecurity protections to commit to a set of best practices, including operating requirements for partners with access to Claude Mythos 5.

After those best-practice commitments were implemented, Anthropic said it restarted external cybersecurity evaluations. The supplied sources do not describe the breach, identify affected systems, quantify damage, specify how long the reassignment will last, or enumerate the complete best-practice framework.