Anthropic says Houthis used Claude AI for missile development programmes

Anthropic has disclosed that the Houthi movement in Yemen utilised its Claude AI system to support work on multiple weapons development programmes, including missile design and testing. The findings were detailed in a report by the San Francisco-based AI safety firm, which documented how the group accessed the tool to advance three separate military initiatives.

The report highlighted that Claude was employed for technical assistance related to weapons systems, with evidence suggesting the group conducted test-firings of rockets developed with aid from the AI platform. The disclosure underscores growing concerns within the technology sector about the potential weaponisation of large language models and generative AI tools by non-state actors and militant organisations.

AnthropiC's findings add to mounting scrutiny of how advanced AI systems are safeguarded against misuse. The company has previously implemented various protective measures to prevent its tools from being used for harmful purposes, though the incident demonstrates vulnerabilities in existing safety frameworks.

The revelation comes amid broader international discussions about AI governance and the responsibilities of technology firms to prevent their systems from supporting violence or weapons development. Security experts have warned that as AI capabilities become more sophisticated and accessible, the risks of misuse by hostile actors increase proportionally.

AnthropiC did not immediately detail what specific measures it would implement in response to the Houthi case, though the report suggested ongoing efforts to refine detection and prevention protocols.