Anthropic says it has disrupted six operations that attempted to develop conventional weapons by using its Claude AI system. The company says these efforts failed and that it later investigated to understand why they did not work.
In its account, Anthropic links the activity to locations including Yemen, Russia, and China, and says the disruptions have occurred since December. The reporting describes an investigation phase in which Anthropic returns to check the outcome of the attempted weapons software use.
Across the available sources, the core claims are consistent: Anthropic identifies multiple weapon-development-related operations, states they were tied to Claude usage, and says it has taken steps to disrupt them while characterizing where and when the attempts were detected. The outlets differ mainly in emphasis on the failed attempt and the subsequent review process rather than on the underlying allegations and geography.
