In contract negotiations in December between senior Pentagon officials and executives from AI giant Anthropic, the company agreed to allow the U.S. government to use its AI systems for missile and cyber defense purposes, said a person familiar with the matter, requesting anonymity to discuss private discussions.
But that clearly did not satisfy the Pentagon.
Following weeks of tension between the Pentagon and Anthropic over restrictions on the military’s use of its products, Defense Secretary Pete Hegseth on Tuesday gave the company’s CEO, Dario Amodei, a strict ultimatum to allow the company to use its AI technology for all legitimate military purposes by Friday or force cooperation, a senior Pentagon official told NBC News.
The ultimatum, detailed to NBC News by a senior Pentagon official, comes as Anthropic, a company that has heavily promoted the safety of AI, seeks to maintain a firm policy of preventing its systems from being used for domestic mass surveillance or direct use in lethal autonomous weapons.
The December contract changes will allow the company’s systems to be widely used in cyber and missile defense, the people said. “Each iteration of our proposed contract language will enable our models to support missile defense and similar applications,” an Anthropic spokesperson told NBC News in a statement.
But the company’s insistence on guardrails continues to be a source of controversy between Anthropic and the Department of Defense.
A Pentagon official said representatives from the department, including Under Secretary of Defense Emil Michael, recently discussed several hypothetical scenarios with Anthropic leaders about how the company’s products might be adopted by the military.
As part of these negotiations, officials discussed how Anthropic’s systems would be used if an adversary launched an intercontinental ballistic missile toward the United States. Pentagon sources said officials discussed whether Anthropic’s guardrails could somehow prevent a U.S. response to the launch. The official said Anthropic officials could be asked to lift those restrictions, but Pentagon leadership was not completely satisfied with Anthropic’s adjustments and did not want to be left in the care of a private company.
An Antropic spokesperson said any suggestion by Amodei that the Pentagon would have to call the company for every missile defense operation was “patently false.”
In the latest escalation of negotiations, Pentagon leaders at a meeting Tuesday said they may invoke the Defense Production Act to force Anthropic to comply with Pentagon rules, according to a senior Pentagon official. The law would allow the president to take control of domestic companies important to national security if necessary.
At Tuesday’s meeting, Pentagon leaders identified Anthropic as a “supply chain risk” and also threatened to ban all defense business with Anthropic unless it agrees terms and conditions for certain high-stakes applications with the Pentagon by Friday, the person said.
“Antropic is required to join the Department of the Army by 5:01 p.m. Friday,” a senior Pentagon official said of the ultimatum in a statement provided to NBC News, responding to questions about the meeting. “If they do not participate in this, the Secretary of the Army will invoke the Defense Production Act against Anthropic to force its use by the Department of Defense.”
“Additionally, the Secretary of the Army will designate Anthropic as a supply chain risk,” the official said.
Asked about Tuesday’s meeting, an Anthropic spokesperson said in a statement: “Mr. Dalio expressed appreciation for the department’s work and thanked the Secretary for his efforts. We continued to discuss usage policies in good faith so that Anthropic can continue to support the government’s national security mission, consistent with what our model can do reliably and responsibly.”
Mr. Hegseth praised Anthropic’s products and said the Pentagon wants to work with the company, according to another person familiar with the meeting. He requested anonymity in order to speak candidly. The official confirmed that the ministry has said it will end Anthropic’s cooperation with the Pentagon by Friday if the terms are not agreed to.
Anthropic’s Claude system was used during an operation to capture Venezuelan President Nicolas Maduro in January, according to reports in the Wall Street Journal and Axios. Exactly how the system was used is unclear.
On January 9, Mr. Hegseth sent a memo to Pentagon leaders announcing that the Pentagon is aiming to become an “AI-first combat force.” He outlined a push to use AI models like Anthropic’s for all legitimate military purposes, “free from the constraints of usage policies” set by individual AI companies.
Anthropic is the only AI company whose products are actively used on classified networks through an agreement with data analytics company Palantir. A senior Pentagon official confirmed to NBC News that xAI reached an agreement with the Pentagon on Monday to use the Grok chatbot system on classified networks, agreeing to allow Hegseth to utilize its system for “any lawful use” as desired.
Anthropic is one of four AI companies (the others being OpenAI, Google DeepMind, and xAI) awarded contracts worth up to $200 million in July to “prototype frontier AI capabilities to advance U.S. national security.”
