President Trump ordered federal agencies to stop using Anthropic’s AI products and the Pentagon moved to designate the company a national security supply‑chain risk, sharply escalating a dispute over military uses of advanced AI.
Hours after the president’s directive, OpenAI announced it had reached an agreement to provide its AI technology for classified Defense Department networks. The developments capped an acrimonious fight between Anthropic and the Pentagon over whether the company could contractually restrict its models from being used for U.S. domestic mass surveillance or to power fully autonomous weapons as part of a potential Pentagon deal worth up to $200 million.
On Truth Social, Mr. Trump denounced Anthropic and directed “EVERY Federal Agency in the United States Government to IMMEDIATELY CEASE all use of Anthropic’s technology,” saying agencies would have a six‑month phaseout. Defense Secretary Pete Hegseth followed by saying he would label Anthropic a supply‑chain risk to national security and blacklist it from working with the U.S. military or its contractors, referring to the department as the “Department of War.” Hegseth said Anthropic could continue services for up to six months to allow a transition to what he called “a better and more patriotic service.”
Anthropic said it would challenge the designation in court, calling it legally unsound and a dangerous precedent for vendors that negotiate use restrictions with the government. The company argued Hegseth lacks statutory authority to bar contractors from using Anthropic’s Claude model for non‑Defense Department customers and said a supply‑chain risk designation should only apply to Defense Department contracts.
Anthropic maintains it negotiated in good faith and supports lawful national security uses of AI, while insisting on two narrow exceptions: no fully autonomous weapons and no models that enable mass domestic surveillance of Americans. The company framed those limits as safety and civil‑rights protections, saying current frontier models are not reliable enough for autonomous weaponization and that mass surveillance would threaten civil liberties.
CEO Dario Amodei defended Anthropic’s stance publicly, arguing military decisions rest with the department and that certain uses exceed what today’s technology can safely do. The Pentagon’s undersecretary for research and engineering, Emil Michael, sharply criticized Amodei on X, accusing him of dishonesty and trying to exert control over military decisions. Pentagon officials say existing law and policy already prohibit domestic mass surveillance and uncontrolled AI weaponization.
The standoff at times featured threats by the government to invoke the Korean War–era Defense Production Act to compel Anthropic to permit certain uses of its technology, while simultaneously threatening blacklisting. Hegseth accused Anthropic of seeking veto power over military operational choices and said the department must have full access to models for all lawful defense purposes.
OpenAI CEO Sam Altman, who had previously voiced reservations similar to Anthropic’s, said OpenAI’s agreement with the Defense Department includes safeguards barring domestic surveillance and requiring human responsibility for use of force—the same “red lines” Anthropic sought. OpenAI, Google and Elon Musk’s xAI already hold Defense Department contracts; xAI was approved earlier in the week for classified settings. OpenAI said it would negotiate exclusions to prevent use for U.S. domestic surveillance or powering autonomous weapons without human approval.
Independent experts called the public clash unusual for Pentagon contracting, where suppliers typically do not dictate how the Defense Department uses products. Analysts said the dispute reflects the novel, untested nature of AI and the rare circumstances in which companies seek to limit government use cases.
The episode also has business implications. Anthropic, described in reports as valued at roughly $380 billion and planning an IPO, generates about $14 billion in revenue; while the potential Pentagon contract was relatively small, the confrontation with the administration could affect investor sentiment and other licensing deals. The showdown highlights a growing fault line over whether AI firms can put contractual limits on government uses of their tools and how to balance national security with safety and civil‑rights concerns as AI becomes integrated into defense systems.
NPR’s Bobby Allyn contributed to this report.