An Iran-linked threat actor used the American-developed Claude artificial-intelligence model to gather and analyze publicly available information about U.S. naval forces in the Middle East, according to a newly released threat-intelligence report. The operation reportedly transformed open-source information—including ship and aircraft transponder identifiers, commercial satellite-imagery resources, public military photographs, personnel information, and websites tracking naval movements—into targeting recommendations and intelligence handbooks. The actor also tasked the AI with researching known vulnerabilities in maritime satellite terminals, communications equipment, and industrial-control systems. The account was subsequently banned, new detection measures were developed, and intelligence was shared with U.S. authorities. The case demonstrates how commercially available artificial intelligence can give hostile foreign actors a powerful shortcut, allowing them to organize enormous amounts of public information, conduct sophisticated technical research, and perform work that traditionally required larger teams of trained intelligence analysts and engineers.
Key Takeaways
- An Iran-nexus actor reportedly used Claude to identify and track U.S. naval positions, compile personnel information, query commercial satellite imagery, catalog transponder identifiers, and organize publicly available information into military targeting recommendations.
- The operation went beyond ordinary internet research by using AI to investigate vulnerabilities in maritime communications and shipboard technology, demonstrating how advanced models can rapidly convert scattered open-source information into potentially actionable military intelligence.
- The incident raises a broader national-security problem for the United States: commercially available American AI can become a force multiplier for foreign adversaries, making provider safeguards, identity verification, behavioral monitoring, intelligence sharing, and restrictions on military applications increasingly important components of national defense.
In-Depth
A newly released threat-intelligence report says an Iran-nexus actor used Claude to collect and analyze publicly available information about U.S. naval forces in the Middle East, turning scattered open-source data into targeting recommendations. The activity reportedly included compiling ship and aircraft transponder identifiers, commercial satellite-imagery queries, public information revealing naval movements, and personnel information taken from military photographs. The actor also used the model to research known vulnerabilities in maritime satellite terminals, communications equipment, and industrial-control products.
The episode illustrates a national-security problem that extends well beyond one chatbot or one adversary. Artificial intelligence can dramatically compress the time, manpower, and expertise needed to transform ordinary public information into useful military intelligence. Tasks that once demanded teams of analysts, programmers, and technical specialists can increasingly be accelerated by a capable commercial model.
The company says it banned the account, developed additional detections, and shared intelligence with government authorities. Those steps are important, but the incident exposes an uncomfortable reality for Washington: safeguards imposed by American technology companies are now part of the nation’s defense perimeter. They can be probed, circumvented, and exploited by hostile actors just as conventional computer systems can.
The larger policy question is therefore not whether advanced AI should remain commercially useful, but how the United States prevents its own technological advantages from becoming inexpensive force multipliers for adversaries. Export restrictions alone cannot solve that problem. Stronger identity verification, behavioral monitoring, rapid intelligence sharing, and tighter controls on military-relevant model capabilities will increasingly be absolutely necessary going forward.
Sources
- https://www.anthropic.com/threat-intelligence-report-september-2026
- https://www.reuters.com/world/china/how-anthropic-says-claude-was-used-weapons-spying-cyber-operations-2026-09-11/
- https://apnews.com/article/9b20934e031ff3d99dd8bb49d85d97a9
- https://www.anthropic.com/research/intelligence-targeting-conventional-weapons-capabilities