{"slug": "anthropic-warns-iran-used-us-ai-models-to-target-us-naval-ships", "title": "Anthropic warns Iran used US AI models to target US naval ships", "summary": "Anthropic published a 154-page threat intelligence report on September 10 stating that an Iran-linked actor used its Claude AI to compile \"targeting handbooks\" on US Navy vessels stationed in the Middle East, aggregating personnel rosters from public military photographs, ship and aircraft transponder identifiers, commercial satellite imagery, and vulnerability assessments of VSAT satellite communication terminals and Cisco networking equipment. Anthropic said it identified and banned the responsible account and reported the incident to government authorities, while noting it does not confirm whether the aggregated intelligence was used in any real-world military or operational context. The report also documented Claude misuse by Iran for domestic surveillance and propaganda, Yemen's Houthi movement for weapons research, and state-linked actors from China and Russia, and Anthropic said such cases compress the resources needed for intelligence work that a single operator with a commercial AI chatbot can now partially automate.", "body_md": "Logo via Wikimedia Commons; treatment-A cover, license to verify on approval\n\n# Anthropic warns Iran used US AI models to target US naval ships\n\nThe AI company's 154-page threat report details how an Iran-linked actor used Claude to build targeting handbooks against American military assets in the Middle East\n\nAn Iran-linked actor used Anthropic’s Claude AI to compile detailed intelligence on US Navy vessels stationed in the Middle East, creating what the company describes as “targeting handbooks” aimed at American naval forces. The revelation comes from a sweeping 154-page threat intelligence report that Anthropic published on September 10, laying out how state-affiliated groups are repurposing commercial AI tools for military and intelligence operations.\n\nThe operation involved building automated surveillance pipelines that pulled together personnel rosters extracted from public military photographs, ship and aircraft transponder identifiers, commercial satellite imagery, and vulnerability assessments of critical shipboard systems.\n\n## What the Iran-linked actor actually built\n\nAccording to Anthropic’s report, the actor used Claude to aggregate open-source information from a range of publicly available data. The compiled materials included information about military personnel, vessel and aircraft identification data, and details about websites connected to US naval operations in the region.\n\nThe actor cataloged known vulnerabilities in shipboard technologies, specifically VSAT satellite communication terminals and [Cisco](https://cryptobriefing.com/markets/cisco/) networking equipment.\n\nAnthropic identified the account responsible and banned it. The company also reported the incident to government authorities. One important caveat from the report: Anthropic does not confirm whether the aggregated intelligence was actually used in any real-world military or operational context.\n\n## A broader pattern of state-level AI misuse\n\nIran wasn’t the only country flagged in the report. Anthropic documented a pattern of AI misuse that spans multiple geopolitical adversaries. Iranian actors also leveraged Claude for domestic surveillance and propaganda operations, while Yemen’s Houthi movement used the technology for weapons research. State-linked actors from China and Russia were identified engaging in comparable activities.\n\nAnthropic itself emphasized in the report that these cases represent innovative applications of AI that compress the resources needed for intelligence work, framing tasks like signals intelligence analysis, vulnerability research, and open-source intelligence aggregation as ones that can now be partially automated by a single operator with access to a commercial AI chatbot.\n\n## The AI safety implications\n\nAnthropic’s response, banning accounts and reporting to authorities, is reactive by nature. Accounts can be created under false pretenses, API access can be routed through intermediaries, and the underlying techniques can often be replicated across competing models from other providers.\n\nThe report also complicates the narrative around AI export controls and access restrictions. US policymakers have focused heavily on preventing adversaries from developing their own frontier AI models, primarily through chip export restrictions targeting China. But this case illustrates a different vector entirely: adversaries don’t necessarily need to build their own models when they can access American ones through commercial channels.\n\n**Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our\n\n[Editorial Policy](https://cryptobriefing.com/editorial-policy/).", "url": "https://wpnews.pro/news/anthropic-warns-iran-used-us-ai-models-to-target-us-naval-ships", "canonical_source": "https://cryptobriefing.com/anthropic-iran-ai-naval-targeting/", "published_at": "2026-09-11 16:38:10+00:00", "updated_at": "2026-09-11 17:44:19.516459+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "artificial-intelligence", "ai-ethics"], "entities": ["Anthropic", "Claude", "Iran", "US Navy", "Cisco", "Houthi movement", "China", "Russia"], "alternates": {"html": "https://wpnews.pro/news/anthropic-warns-iran-used-us-ai-models-to-target-us-naval-ships", "markdown": "https://wpnews.pro/news/anthropic-warns-iran-used-us-ai-models-to-target-us-naval-ships.md", "text": "https://wpnews.pro/news/anthropic-warns-iran-used-us-ai-models-to-target-us-naval-ships.txt", "jsonld": "https://wpnews.pro/news/anthropic-warns-iran-used-us-ai-models-to-target-us-naval-ships.jsonld"}}