Anthropic reports Iranian-linked actors used Claude AI for missile and reconnaissance efforts

Anthropic has released a report detailing how threat actors linked to Iran have weaponized its Claude AI models for military and reconnaissance purposes. The company identified a weapons development cell in northern Yemen that used Claude Code to assist in three distinct programs: a guided rocket, a multi-stage ballistic missile with a range exceeding 2,000 kilometers, and a missile system involving a hypersonic glide vehicle.

In these instances, the actors utilized the AI as a coding partner to develop guidance, navigation, and control software, performing tasks such as writing code, researching technical problems, and running simulations. While Anthropic reported that a test-fire of a guided rocket appeared to fail and no operational weapons have been successfully fielded, the company has since blocked the complicit accounts.

Beyond missile development, the report highlights an “Iran-nexus” using the platform for military reconnaissance against United States personnel. This included using Claude to analyze publicly available data for targeting recommendations, creating scripts to track naval ships via satellite imagery and transponder identifiers, and developing Python scripts to identify American personnel from military photographs. The AI was also used to research vulnerabilities in naval electronic systems.