Unsealed court filings reveal how Anthropic's Pentagon talks collapsed over AI safeguards

A series of newly unsealed court filings has revealed how negotiations between artificial intelligence company Anthropic and the US Department of Defense (DoD) broke down over the military's use of AI, exposing deep disagreements on autonomous weapons and surveillance, as per a Wall Street Journal report.The documents, released as part of Anthropic's ongoing legal challenge against the Pentagon, include email exchanges between Anthropic CEO Dario Amodei and Emil Michael, the US Undersecretary of Defense for Research and Engineering. The correspondence outlines months of negotiations that ultimately ended with the company being designated a "supply-chain risk" by the Pentagon.According to the filings, Anthropic sought contractual safeguards that would prohibit its AI models from being used for fully autonomous weapons systems and domestic mass surveillance. Amodei argued these restrictions were necessary to establish clear limits on the deployment of advanced AI while allowing legitimate national security and foreign intelligence applications.The Pentagon, however, pushed for broader contractual language that would permit the technology to be used for all lawful purposes. The department maintained that operational requirements did not allow for the distinctions Anthropic was proposing, particularly between offensive and defensive military applications.Also read: TCS, Infosys, Wipro and HCLTech likely to begin FY27 with muted quarter as AI, weak demand weigh on growthThe disagreement intensified through January and February, with both sides exchanging revised proposals but failing to reach common ground. In one of the final exchanges, Amodei rejected a draft agreement, saying it effectively removed the safeguards Anthropic considered essential.The following day, then-US Defense Secretary Pete Hegseth announced that Anthropic had been designated a supply-chain risk, restricting Pentagon partners from using the company's technology.Anthropic subsequently challenged the decision in court, arguing the designation amounted to retaliation for its public positions on AI governance and national security. While a federal judge initially granted preliminary relief to the company, an appeals court later reversed that order. The case remains ongoing.The filings also include references to internal government discussions after the designation was announced, as agencies reviewed contracts involving Anthropic's technology.