AI developer Anthropic disclosed on Thursday that Iran-linked accounts leveraged its Claude artificial intelligence model for various malicious activities, including propaganda, domestic surveillance, and targeting US naval forces. The report, covering activities between December 2025 and August 2026, detailed that one Iran-linked account utilized Claude to compile targeting handbooks and track naval positions using publicly available information. This included collecting names of US personnel from public military photographs, ship and aircraft transponder identifiers, commercial satellite-imagery query scripts, and lists of websites exposing US naval movements.
Furthermore, the same account used Claude to research vulnerabilities in shipboard systems, specifically mentioning maritime satellite terminals, Cisco communications equipment, and industrial control products. It also analyzed known Common Vulnerabilities and Exposures (CVEs) in maritime VSAT terminals. Separately, the AI was employed to develop components for an Iranian state domestic surveillance platform, combining automatic license-plate recognition with mobile-device identifier interception, and to map social connections within a private Telegram group.
Anthropic stated it banned the accounts involved and shared threat intelligence with government authorities and industry partners, though it remains unclear if the intelligence generated was used in any specific Iranian military operation. This disclosure follows recent incidents where Iranian forces targeted US naval assets and facilities in the Gulf region, though US officials reported no casualties. The news also comes amidst concerns about the depletion of US Patriot missile inventories, with estimates suggesting about 65% of the pre-war stock, or around 1,500 Patriot interceptors, has been expended in the conflict with Iran.
Separately, other reports indicate that Anthropic's Claude AI was also misused by users from countries like Russia, China, and Yemen for a wide range of potential military applications, including kamikaze drone swarms, missile navigation systems, and biological weapons research. Iranian state-aligned actors also used Claude to gather open-source intelligence and construct targeting recommendations against US naval forces operating in regional waters, as well as to write guidance software for long-range ballistic missiles. Goldman Sachs warned that escalating attacks in the Persian Gulf and Red Sea could drive Brent crude prices above $120 a barrel, while Reuters reported that Tehran routed an estimated $2 billion to $2.5 billion through a China-based trade mechanism over the past year.
