Anthropic Bans Iranian Accounts That Used Claude To Target US Navy
Anthropic has banned an Iran linked account that was using its Claude models to compile intelligence and targeting material aimed at U.S. naval ships in the Middle East.
In a 154 page intelligence report published in September 2026, titled “Detecting and Countering Misuse of AI,” Anthropic describes an Iranian group that used Claude to identify and track positions of American naval ships across the region.
According to Anthropic, the Iranian group used its Claude models to build an automated system to collect public military photographs, commercial satellite images, and data from public websites that track naval movements. The system also collected details of U.S. personnel from public military photographs, including ship and aircraft particulars. The same account has also been used to find vulnerabilities in shipboard systems and maritime communication equipment.
Anthropic added that the account was linked to an Iranian domestic surveillance project that dealt with automatic license plate recognition. AI models also analysed data from a leaked Telegram group of 244 members and mapped their social connections.
The intelligence report, however, does not name the individuals or organisations behind the account. It also doesn’t specify naval vessels and operating bases. Anthropic describes the activity only as directed against US naval forces in the Middle East. The company said it could not confirm whether the compiled material was later used to plan or support any operation against US forces. Anthropic banned the account and shared its findings with government authorities.
The report also describes that Anthropic’s Claude, Haiku, Sonnet, and Open Opus models have been exploited for cyber operations, surveillance, influence campaigns, and weapon development tied to Iran, Russia, and China.
Three more Iran-linked accounts have been identified as being linked to government propaganda bodies associated with the Islamic Organisation under Iran’s Ministry of Culture and Islamic Guidance. These accounts used Claude to build influence campaigns aimed at audiences inside and outside the country.
Anthropic said it used each case to enhance its security and shared the technical details with government and industry partners. The report’s data aims to help other AI developers recognise similar patterns of misuse in military and intelligence operations.
Reference: Reuters/WSJ
RELATED POSTS
Disclaimer :
The information on this website is for general purposes only. While efforts are made to ensure accuracy, we make no warranties of any kind regarding completeness, reliability, or suitability. Any reliance you place on such information is at your own risk. We are not liable for any loss or damage arising from the use of this website.
Disclaimer :
The information on this website is for general purposes only. While efforts are made to ensure accuracy, we make no warranties of any kind regarding completeness, reliability, or suitability. Any reliance you place on such information is at your own risk. We are not liable for any loss or damage arising from the use of this website.
Related Articles
Subscribe To Our Daily Newsletter
By subscribing, you agree to our Privacy Policy you can unsubscribe anytime.
BE THE FIRST TO COMMENT