Reports from international media outlets indicate that the United States Department of Defense deployed Anthropic’s Claude AI model during large-scale strikes on Iran, despite a presidential directive to halt federal use of the technology. The strikes, part of a coordinated operation with Israel, took place on February 28, 2026, and were publicly reported as involving the use of Claude for tasks including intelligence assessments, target selection and battlefield simulations.
The controversy stems from a policy decision by President Donald Trump and Pete Hegseth, who designated Anthropic a “supply chain risk” and instructed all federal agencies to stop using its AI systems. The move, part of an escalating dispute between the administration and the AI company, was widely covered in the news and contrasted with the military’s continued use of Claude in active operations.
According to reports, Claude’s continued deployment reflects its deep integration across US military networks, particularly within United States Central Command infrastructure. At the time of the strikes, Claude remained embedded in systems used for intelligence and operational support. Because of this integration, the complete withdrawal of the AI system was not immediately feasible. Contractual terms reportedly allow for a phased transition, with the Pentagon giving authorities and partners several months to offboard Claude and transition to alternative platforms.
The timing of the strikes and the use of Claude generated discussion about the complexities of balancing national security directives with operational requirements. Experts noted that stopping use of a system deeply woven into classified networks often requires careful planning and technical preparation. In this context, Claude’s presence in the operation underscored how entrenched AI tools have become in modern military planning and real-time decision processes.
Anthropic has maintained that it supports lawful uses of its AI models for national security, while insisting on safeguards against applications it views as ethically problematic, such as mass domestic surveillance and fully autonomous weapons. These ethical principles were at the centre of negotiations with the Pentagon and contributed to the dispute that led to the “supply chain risk” designation. The company has stated it will challenge that label in court, arguing it is unprecedented for an American AI firm.
In the midst of the controversy, the Pentagon has also moved forward with arrangements involving other AI providers. For example, OpenAI reached agreements to provide AI tools for certain classified military needs, reflecting a shift in the defence sector’s approach to advanced artificial intelligence integration.
The situation highlights ongoing tensions between technology providers and government directives, particularly where ethical constraints and national security demands intersect. As the transition away from Claude unfolds over the coming months, defence officials have indicated they will continue to balance operational readiness with evolving policy requirements and contractual arrangements.
