Iran and Houthis Leveraged AI for Espionage, Weapons Development

Iran and its Houthi allies have been using the AI model Claude for intelligence gathering, surveillance of Israelis, identifying military targets, and developing missile technology, according to an investigation by Anthropic, the company behind Claude. The findings detail how Iran and affiliated groups exploited the AI for intelligence, military, and repressive purposes between late 2025 and August 2026.

One user linked to Iran's Islamic Revolutionary Guard Corps created an automated intelligence tool using Claude. This system collected and analyzed open-source information on hundreds of Israelis and members of Jewish communities in Western countries, generating profiles of Israeli and American citizens, including government officials and NGO staff, to supplement target lists. Another Iranian user employed Claude for military applications, tracking U.S. Navy vessels, identifying military personnel, and assessing vulnerabilities in maritime infrastructure. Anthropic noted that while AI is expanding its capabilities beyond analysis to complex operations, human decision-making remained crucial for target selection.

The Houthi group in Yemen utilized Claude as a substitute for software engineers in their efforts to develop advanced missile technology, including a ballistic missile with a range exceeding 2,000 kilometers and a hypersonic glide vehicle. Following a failed test, engineers used the AI to analyze the cause of the failure. Anthropic found no evidence that these developments reached operational status.

Within Iran, Claude was also used for domestic surveillance. Two security units created 16 accounts to monitor social media posts and profile thousands of citizens. One unit developed a malicious browser extension disguised as a prayer schedule app to covertly collect data from users of Iranian social networks. Iranian propagandists also used the AI to create fake personas and networks for disseminating messages, attributing false claims to Western research centers to enhance credibility.

Anthropic also identified the use of Claude by the Mujahedin-e Khalq organization to mimic a real activist based on their online publications. Additionally, entities linked to the UAE government used the model to prepare false testimonies for UN institutions. An Israeli-Singaporean company experimented with the platform to analyze users in Iran and the Persian Gulf, aiming to determine political leanings and create fake profiles, but Anthropic blocked access before its deployment.