Artificial intelligence company Anthropic reported that an Iran-linked group used its Claude AI model to attempt to target U.S. naval forces. According to the company, the unidentified group, which it labeled a "threat actor," utilized the AI to aggregate open-source information, identify naval positions, and generate targeting recommendations. Anthropic stated it has since banned the account and shared the findings with the U.S. government to disrupt the activity.
The report details how the group analyzed publicly available data, including satellite imagery, ship and aircraft tracking information, and captions from military photographs, to build a picture of U.S. movements. While Claude is officially blocked in Iran, researchers noted that individuals can bypass these restrictions using virtual private networks (VPNs) and foreign phone numbers. The Pentagon declined to comment on the report, stating it does not discuss intelligence matters.
In addition to military targeting, Anthropic identified three Iranian state-aligned accounts using Claude for public influence campaigns. These accounts reportedly generated propaganda that was falsely attributed to research institutions like the Brookings Institution and the RAND Corporation to build credibility. The content was distributed across social media platforms, including X, TikTok, and Instagram, to shape the information landscape regarding conflicts involving the U.S. and Israel.
This development affects U.S. military personnel, particularly naval forces, by potentially reducing the effectiveness of traditional methods used to obscure troop and vessel movements. According to Sam Bresnick, a research fellow at Georgetown’s Center for Security and Emerging Technology, advanced AI models are making it easier for adversaries to track the U.S. military, which had previously maintained an advantage in situational awareness. The scale of this issue extends beyond Iran; a July report from Recorded Future indicated that AI has likely accelerated Iran's cyber and military capabilities, potentially aided by tactics refined by Russia during the conflict in Ukraine.
For the average service member or their family, this shift could eventually change day-to-day operational security protocols and how information is shared publicly. Nikita Shah of the Center for Strategic and International Studies (CSIS) noted that AI allows adversaries to carry out attack methods more efficiently and at a faster pace than traditional manual data analysis. While the exact number of individuals targeted was not reported, Shah stated the use of AI to refine missile and drone targeting directly impacts the safety of personnel.
The findings suggest that American-developed commercial technology is being used to support the strategic objectives of foreign adversaries. This has led to new countermeasures; one proposal reviewed by CBS News involves the Pentagon creating AI "personas" to mimic realistic patterns of life and generate "digital exhaust" to confuse adversarial surveillance. As these models become more capable and accessible, the national security community expects a continuous cycle of technological refinement and countermeasure development between the U.S. and its adversaries. No specific dates for the implementation of these new military countermeasures were reported.