Iranian operatives used the publicly available Claude artificial intelligence platform to collect targeting information on American service members and US Navy forces operating across the Middle East, according to a new report from Anthropic.
The disclosure offers another warning that hostile regimes are turning commercial AI tools into digital reconnaissance assistants.
Anthropic said Thursday that the activity occurred between December 2025 and August 2026.
During that period, the Iranian threat actor allegedly relied on Claude to analyze public information and recommend methods for locating American naval assets throughout the region.
Here's What They're Not Telling You About Your Retirement
MORE NEWS: A Quarter Century Later, America Mourns September 11th with Unfading Grief and Solemn Reflection
“The threat actor used Claude to compile targeting handbooks…to identify and track naval positions based on open-source information,” the report said.
That work reportedly transformed scattered public records into organized material that could be used to monitor ships, aircraft and individual service members.
The operatives gathered names and other information about US military personnel from publicly released military photographs.
While such images may appear routine, hostile intelligence collectors can combine faces, names, unit details and locations to build a far more revealing operational picture.
This Could Be the Most Important Video Gun Owners Watch All Year
Claude was also used to examine ship and aircraft transponder data, which broadcasts identifying codes and location information.
That data can help civilian operators avoid collisions, but in the hands of an Iranian threat actor, it can also become a convenient method for following American movements.
The account additionally sought commercial satellite imagery query scripts designed to locate precise geographic visuals.
Those scripts could make it easier to search vast imagery databases for ports, airfields, ships and other military activity without requiring the resources of a traditional state intelligence operation.
Anthropic said the Iranian actor assembled information about public websites that track US Navy movements.
Individually, those sources may seem harmless, but combined through an AI platform, they can provide hostile forces with a faster and more organized targeting process.
The threat actor did not stop at locating American forces. According to the report, it also used Claude to research weaknesses in US Navy shipboard systems that could potentially be exploited, raising concerns about cyber intrusion, electronic warfare and physical attacks.
Anthropic said it banned the account after discovering the activity.
The company also said it established additional safeguards intended to block similar misuse and shared relevant information with government agencies responsible for addressing the threat.
The report described other troubling cases involving Claude, including a group of Yemeni operatives that allegedly used the platform to develop targeting software for guided missiles.
That episode demonstrates how commercial AI can reduce technical barriers for terrorist groups and regional proxies seeking more accurate weapons.
Internal concerns surrounding the company have also intensified. Jacob Coxon, a researcher associated with Anthropic and OpenAI, resigned Tuesday and questioned the conduct of both companies in a lengthy post on X.
“Neither company is acting responsibly,” said Coxon. “They are racing straight to self-improving superintelligence and gambling with our lives.”
The US military previously used Anthropic technology but ended its relationship with the company after Anthropic refused to accept War Department terms governing military access to Claude.
Anthropic argued that those conditions could permit the platform to be used for mass surveillance or fully autonomous weapons.
That clash highlights a growing national security dilemma because American forces need advanced AI capabilities, but private technology companies increasingly want to dictate how elected leaders and military commanders may use them.
Meanwhile, adversaries such as Iran appear perfectly willing to exploit the same tools without pausing for Silicon Valley ethics seminars.
The Iranian activity shows that publicly available information can become dangerous when artificial intelligence rapidly collects, sorts and converts it into usable targeting guidance.
Protecting American sailors and service members will require tighter operational security, stronger counterintelligence and serious scrutiny of how AI companies police hostile foreign accounts.
For the Navy, this is not an abstract technology debate taking place inside a corporate conference room.
It is a direct force protection problem involving real ships, real crews and an Iranian regime that has repeatedly used proxies, missiles and surveillance to threaten American personnel in the Middle East.
Join the Discussion
COMMENTS POLICY: We have no tolerance for messages of violence, racism, vulgarity, obscenity or other such discourteous behavior. Thank you for contributing to a respectful and useful online dialogue.