Published 5 hours ago • loading... • Updated 30 minutes ago
Iran Used Claude to Target US Navy in Middle East, Anthropic Says
Anthropic said it banned the account and shared the case with authorities after the actor assembled targeting data on U.S. Navy warships.
On Thursday, San Francisco-based Anthropic released a report detailing how an Iran-linked actor exploited its AI assistant, Claude, to compile "targeting books" for tracking U.S. Navy warships in the Middle East.
Exploiting Claude, the actor scraped U.S. personnel names from public military photos, collected ship transponder identifiers, and researched software vulnerabilities in maritime satellite communications and Cisco equipment.
Researcher Jacob Coxon resigned Tuesday, criticizing Anthropic and OpenAI's ethical approaches to AI development, as the company acknowledges that while safeguards intercepted significant misuse, some malicious requests succeeded.
Anthropic also detailed state-linked misuse, including a China-based actor preparing briefings on U.S. anti-torpedo systems and a cell in Yemen developing guidance software for ballistic missiles.
Following an Aug. 19 message from Navy Secretary Hung Cao urging personnel to heighten vigilance against security threats, Anthropic banned the account linked to the activity.