A threat intelligence report has identified a cell operating out of northern Yemen that allegedly attempted to use Anthropic’s Claude artificial intelligence model to assist in the development of ballistic missile technology. According to the report, the group sought to leverage the AI to create guidance, navigation, and control software for three separate weapons development programs.

While the report does not explicitly name the Houthi movement as the direct operator, it notes that the activity occurred within territory primarily controlled by the group. The findings, which were first reported by the Financial Times, underscore a growing concern among security analysts regarding how generative AI tools might be repurposed by non-state actors or militant groups to advance their military capabilities.

Anthropic has not publicly confirmed the specific details of the interaction, but the report serves as a case study for the risks associated with large language models. The incident highlights the challenges tech companies face in implementing safety guardrails to prevent their platforms from being used for illicit or dangerous activities. As of now, there is no evidence suggesting the group successfully deployed the software in a functional weapon system, but the attempt itself marks a significant development in the intersection of AI and global security threats.