Anthropic has walked away from its deal with the Pentagon after learning that the U.S. military wanted to use the company's AI to analyze bulk data collected from American citizens, The Atlantic has revealed. The data in question could include chatbot queries, Google search histories, GPS-tracked movements, and credit-card transactions - all of which could be cross-referenced with other personal details.
As reported by "Hvylya", the new details come from The Atlantic, citing a source familiar with the negotiations who was granted anonymity because they were not authorized to discuss the talks.
According to the report, on Friday morning Anthropic received word that Hegseth's team would make a major concession by removing loophole language from the proposed agreement. The Pentagon had pledged not to use the AI for mass domestic surveillance or fully autonomous killing machines, but qualified those pledges with phrases like "as appropriate" - suggesting the terms could shift based on the administration's interpretation.
Anthropic's team was reportedly relieved by the concession on loophole language. But the relief was short-lived: on Friday afternoon, the company learned that the Pentagon still insisted on using its AI to analyze bulk data collected from Americans. Anthropic's leadership told Hegseth's team "that was a bridge too far, and the deal fell apart," The Atlantic reports.
Soon after the collapse, Defense Secretary Pete Hegseth directed the U.S. military's contractors, suppliers, and partners to stop doing business with Anthropic. The list of affected companies is extensive and includes Amazon, which supplies much of Anthropic's computing infrastructure. Anthropic's AI model is currently the only one allowed into the federal government's classified systems.
The Department of Defense did not respond to The Atlantic's request for comment. Anthropic referred the outlet to its public statement addressing Hegseth's remarks.
Also read: Anthropic Ban May Reshape AI Industry More Than Iran War, MIT Economist Warns