Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei a Friday deadline to resolve a dispute over access to the company’s AI model, Claude, or risk losing a $200 million Pentagon contract, according to multiple reports.
The standoff centers on the Pentagon’s demand for unrestricted access to Claude for “all lawful use,” while Anthropic seeks guarantees that the AI will not be used for mass surveillance of American citizens or to control autonomous weapons systems. Sources familiar with the discussions, as reported by CNN and CBS News, say the Pentagon is prepared to terminate the contract if an agreement isn’t reached by 5:01 p.m. On Friday.
Hegseth, in a meeting with Amodei on Tuesday at the Pentagon, likewise threatened to invoke the Defense Production Act, compelling Anthropic to operate with the military regardless of the company’s objections, and to label Anthropic a supply chain risk, CNN reported. The Defense Production Act, originally used during the Trump administration to address pandemic-related supply shortages, allows the government to influence businesses in the interest of national defense.
Anthropic has expressed concerns about the reliability of AI in high-stakes military applications, specifically citing the potential for “hallucinations” and unintended consequences in targeting decisions, CBS News reported. Amodei reportedly wants assurances that Claude will not be used for final targeting decisions without human oversight. The company also maintains that current laws and regulations do not adequately address the ethical and legal implications of using AI for mass surveillance.
Pentagon officials, however, have dismissed these concerns, asserting that the military’s requests are limited to lawful activities. A senior Pentagon official told CNN that the dispute “has nothing to do with mass surveillance and autonomous weapons being used.”
The conflict comes as the Pentagon seeks to integrate AI capabilities into its operations. Anthropic was awarded the $200 million contract in July to develop these capabilities, according to CBS News. The company is currently the only AI provider with its models deployed on the Pentagon’s classified networks and providing customized models to national security customers.
Other AI companies, including xAI, owned by Elon Musk, are reportedly more willing to cooperate with the Pentagon’s demands. A Pentagon official indicated that xAI’s Grok model is already being used in a classified setting, and that other companies are nearing similar agreements, according to CNN. Negotiations between Anthropic and the DOD have reportedly become increasingly strained in recent weeks, as reported by CNBC.
The meeting between Hegseth and Amodei followed reports from Axios detailing the escalating tensions. As of Wednesday, no resolution had been publicly announced.
