The Anthropic military ban issued by President Donald Trump in late January did not stop U.S. forces from deploying the company’s Claude AI during strikes on Iran just weeks later. On February 28, American and Israeli forces launched “Operation Epic Fury,” targeting Iranian nuclear facilities and military infrastructure – and Pentagon officials confirmed Claude was active throughout.
The AI tool supported intelligence assessments, target selection and battlefield simulations during the operation. Its continued use came down to a single practical reality: Claude was already embedded inside classified and military systems before the ban took effect. Federal officials granted a transition window of roughly six months to phase it out.
The underlying dispute involves two hard limits Anthropic refuses to cross:
- Enabling domestic mass surveillance
- Supporting fully autonomous weapons
Defense Secretary Pete Hegseth publicly demanded unrestricted model access for all lawful military purposes. Anthropic refused and said it will challenge its “supply chain risk” designation in court, calling the label an attempt at “intimidation.”
The ban applies to Department of Defense contracts. Other government and commercial operations continue unaffected.
Following the break, OpenAI reached a separate agreement with the Pentagon, granting ChatGPT access to classified networks.
