The head of artificial intelligence firm Anthropic has rejected a request from the Pentagon to remove certain safety safeguards from its AI systems. The decision follows pressure from US Defence Secretary Pete Hegseth.

Mr Hegseth had previously warned that the company could be removed from the Department of Defense supply chain if it did not comply. The request focused on changes to built-in safety measures designed to limit how the company’s AI tools are used.

Anthropic has declined to alter those safeguards. The company maintains that its safety controls are a core part of its technology. The disagreement highlights tensions between technology firms and the US military over how AI systems should be developed and deployed. Further details about the specific safeguards involved have not been made public.