Anthropic says it disrupted the operation, which compiled ship transponder data, military photos, and satellite imagery to track American vessels
Anthropic says an Iran-linked threat actor used its Claude AI model to compile targeting handbooks on U.S. Navy warships operating in the Middle East, according to a threat intelligence report the company published this week.
The actor used Claude, with assistance from a Python pipeline Claude helped build, to collect and analyze publicly accessible data on U.S. naval forces. The compiled material included a roster of U.S. personnel scraped from captions on public military photographs, ship and aircraft transponder identifiers, commercial satellite-imagery query scripts, and an inventory of public websites that exposed U.S. naval movements, the report said. The actor also directed Claude to research vulnerabilities in shipboard systems, cataloging known software flaws in maritime satellite communications terminals, Cisco $CSCO communications equipment, and industrial control products.
Anthropic said it banned the account associated with the activity, developed new detections, and shared threat intelligence with government authorities.
The naval targeting operation appeared in a section of the report also covering a separate Iran-linked actor that used Claude to build an automated identity-profiling tool targeting hundreds of Israeli government and non-government individuals, and to modify open-source credential-theft malware. A third Iran-linked case described in the same section involved building domestic surveillance software for use against Iranians inside the country.
The 154-page report covers activity Anthropic detected and disrupted between December 2025 and August 2026 across seven categories: cyber operations, influence operations, surveillance, scams and fraud, biological misuse, conventional weapons development, and illicit distillation of its AI models.
The report details a range of state-linked misuse. A suspected Russian state-linked group consistent with public reporting on the actor known as Midnight Blizzard used Claude to automate cyberattacks against Ukrainian government targets and drone manufacturers. Chinese government-aligned actors used the AI to surveil Uyghur diaspora communities in Syria and build dossiers on religious leaders across Asia. A cell in northern Yemen used Claude to develop guidance software for a multistage ballistic missile and test-fired a guided rocket, returning to Claude within hours to analyze the failure, according to the report.
Anthropic acknowledged that its safeguards intercepted a significant portion of the misuse attempts it documented, though some requests succeeded in getting through. Threat actors used techniques including hiding their goals, fragmenting requests across multiple sessions, and routing traffic through virtual private networks to circumvent geographic access controls.
"We really do earnestly believe AI could kill all humans!" Evan Hubinger, a scientist at Anthropic, wrote in a social-media post cited by the Wall Street Journal , stating further that in his view Anthropic lacked a viable path to solving alignment before the arrival of superintelligent systems.
Anthropic said it shared findings from each investigation with authorities and industry partners and used the results to strengthen its detection systems.
The essential business news, delivered fresh every morning.
Join 500,000+ readers who start their day with Quartz.
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
