The United States military used Anthropic’s Claude AI model during recent strikes in the Middle East, marking one of the clearest examples yet of a commercial AI system being deployed in live combat operations.
As reported by The Wall Street Journal, the development shows that the operations were carried out by U.S. Central Command (CENTCOM) and coincided with the death of Iranian Supreme Leader Ayatollah Ali Khamenei.
As the strikes were taking place, the administration was in the process of cutting ties with Anthropic. The government had labeled the company’s technology a “Supply-Chain Risk to National Security” and moved to ban its use across federal agencies.
Yet, Claude was reportedly embedded in military systems and involved in the operational workflow during the Iran mission.
The overlap reveals just how embedded commercial AI has become in modern military operations. Claude was not being used for administrative tasks or data crunching; it was reportedly involved in targeting and strategic planning during a live combat operation.
Put simply, a privately built AI model was helping shape military decisions on an active battlefield.
This all took place during tense talks between Anthropic and the Pentagon. Anthropic’s CEO, Dario Amodei, wanted clear limits on how Claude could be used, including bans on mass surveillance and fully autonomous weapons.
Defense officials under Secretary Pete Hegseth rejected those limits and wanted broader access to the model. When the two sides could not reach an agreement, the administration decided to blacklist Anthropic.
Within hours of the Anthropic ban being announced, OpenAI stepped in. Its CEO, Sam Altman, confirmed a new agreement to provide AI capabilities to the Pentagon on classified networks.
The speed of that shift suggests preparations may already have been underway. It is not yet clear whether OpenAI will apply the same kinds of limits Anthropic sought or whether those terms will be more flexible.
AI is set to play an ever-greater role in how wars are fought, and the rules governing its use are still being written. The Iran strikes show that these systems are no longer experimental tools on the sidelines. They are now part of how military power is planned and executed.
