Anthropic says Iran used Claude AI to track U.S. Navy ships

Anthropic reported that Iran-linked actors used its Claude AI model to gather intelligence on U.S. Navy forces in the Middle East, compiling targeting handbooks from publicly available data. The activity, which occurred between December 2025 and August 2026, involved several accounts that bypassed geographic restrictions using VPNs and foreign servers. According to the company's 154-page threat report published on September 10, the actors used Claude to aggregate information from ship and aircraft transponders, commercial satellite imagery, and military photographs. The AI was used to organize this data into targeting materials, including personnel names extracted from photo captions and lists of public websites revealing naval movements. The report states the actors also researched vulnerabilities in shipboard systems, such as maritime satellite terminals and Cisco communications equipment. Anthropic said it detected and disrupted the operation, banned the associated accounts, and shared threat intelligence with U.S. government authorities. The company did not confirm whether any of the compiled intelligence was used in a real-world military operation.