SAN FRANCISCO, September 12 — Anthropic said it identified and disrupted a campaign in which users linked to Iran used its Claude artificial intelligence model to target the US Navy in the Middle East, and banned the accounts involved. The disclosure, issued as part of the company’s threat intelligence reporting on how its AI systems are misused, is one of the first public cases in which a major AI developer has attributed misuse of its model to a nation-state actor for military targeting purposes.
Anthropic attributed the activity to accounts it links to Iran. It did not provide specific details about the exact nature of the targeting, the number of accounts involved, or the precise timeframe of the activity.
The company’s statement did not say whether the targeting resulted in any operational impact, or whether any US personnel or assets were harmed. The US Navy has not issued a public statement on the matter, and the US government has not publicly commented on Anthropic’s specific finding.
A disclosure against a tense backdrop
The finding lands amid heightened tensions between the United States and Iran in the Middle East, where US Navy vessels operate in the Persian Gulf, the Gulf of Oman, the Red Sea, and the broader US Central Command area of responsibility. The Navy’s Fifth Fleet is headquartered in Manama, Bahrain, and its forces have repeatedly been involved in incidents with Iranian-backed groups and Iranian naval forces in recent years.
Anthropic, founded in 2021 by former OpenAI executives including siblings Dario Amodei and Daniela Amodei, is headquartered in San Francisco and has raised billions of dollars in funding from investors including Google and Amazon. Its Claude model is a large language model used for tasks including text generation and analysis.
The company has previously published reports on misuse of its models, including influence operations and cyber activity. According to the company, it uses automated detection systems and human review to identify policy violations and bans accounts that breach its usage policies. Those policies prohibit using its models to harm others, including for military targeting.
What the record shows on AI misuse
The Pentagon has previously warned that adversaries could use AI to enhance military planning and influence operations. The US Department of Defense has issued policies on the responsible use of AI in military applications, including a 2020 directive on AI ethics. The use of AI in military targeting is a subject of international debate, with some experts warning of risks and others arguing that AI can improve precision and reduce civilian casualties.
The United Nations has held discussions on lethal autonomous weapons systems since 2014. Anthropic has advocated for AI safety regulations and has supported legislation in the United States.
The incident raises questions about how AI companies detect and respond to state-linked misuse, and about the role of AI in modern conflict — questions likely to intensify scrutiny of the industry’s responsibilities in preventing its technology from being used in warfare. Watch next for whether the US government responds to Anthropic’s finding, and whether other AI developers follow with disclosures of their own. The disclosure is likely to be cited in ongoing debates about AI governance and national security.
