Anthropic says it identified a cell of threat actors based in northern Yemen that attempted to use its Claude AI model to support weapons development. Multiple outlets report Anthropic disrupted illegal Claude use tied to three weapons development programmes run by the group.
The reports describe the effort as focused on guided weapons, including missile and rocket systems. SecurityWeek adds that while the users did not succeed in “fielding an operational device,” they carried out a failed test of a guided rocket. Bloomberg and other outlets link the activity to territory controlled by Iran-backed Houthi militants, while the specific operational outcome beyond the failed test is not presented in detail by all sources.
Outlets largely align on Anthropic’s role and location (northern Yemen), but differ in emphasis: some frame the activity as “seeking guided weapons,” others as attempts to “develop advanced missiles,” and SecurityWeek underscores the lack of operational deployment. The reporting collectively highlights concerns about how advanced AI tools could be used by non-state actors for military-related research and testing.
