For months, the Department of Defense and the artificial intelligence company Anthropic have been negotiating a contract over the use of A.I. on classified systems by the Pentagon.
This week, those discussions erupted in a war of words.
On Monday, a person close to Defense Secretary Pete Hegseth told Axios that the Pentagon was “close” to declaring the start-up a “supply chain risk,” a move that would sever ties between the company and the U.S. military. Anthropic was caught off guard and internally scrambled to pinpoint what had set off the department, two people with knowledge of the company said.
At the heart of the fight is how A.I. will be used in future battlefields. Anthropic told defense officials that it did not want its A.I. used for mass surveillance of Americans or deployed in autonomous weapons that had no humans in the loop, two people involved in the discussions said.
But Mr. Hegseth and others in the Pentagon were furious that Anthropic would resist the military’s using A.I. as it saw fit, current and former officials briefed on the discussions said. As tensions escalated, the Department of Defense accused the San Francisco-based company of catering to an elite, liberal work force by demanding additional protections.
The disagreement underlines how political the issue of A.I. has become in the Trump administration. President Trump and his advisers want to expand technology’s use, reducing export restrictions on A.I. chips and criticizing state regulations that could be perceived as inhibitors to A.I. development. But Anthropic’s chief executive, Dario Amodei, has long said A.I. needs strict limits around it to prevent it from potentially wrecking the world.
Emelia Probasco, a senior fellow at Georgetown’s Center for Security and Emerging Technology, said it was important that the relationship between the Pentagon and Anthropic not be doomed.
“There are war fighters using Anthropic for good and legitimate purposes, and ripping this out of their hands seems like a total disservice,” she said. “What the nation needs is both sides at the table discussing what can we do with this technology to make us safer.”
In a statement, Sean Parnell, the Pentagon spokesman, said the department’s relationship with Anthropic was “being reviewed.”
“Our nation requires that our partners be willing to help our war fighters win in any fight,” he said. “Ultimately, this is about our troops and the safety of the American people.”
Anthropic said it was committed to using its A.I. to support U.S. national security and was “having productive conversations, in good faith, with the Department of War on how to continue that work and get these complex issues right.”
Trump Administration: Live Updates
Updated
- Donald J. Trump International Airport? The president’s company trademarked it.
- Workplace inspections by OSHA dropped over a six-month period of 2025.
- Retail billionaire says Epstein ‘conned’ him, but Democrats are skeptical.
A Pentagon official praised the other A.I. companies that have been working on the Pentagon’s unclassified systems. The official said the department had an agreement in place for one of the firms to begin work on the classified system, replacing Anthropic.
The Defense Department has used Anthropic’s technology for more than a year as part of a $200 million A.I. pilot program to analyze imagery and other intelligence data and conduct research. Google, OpenAI and Elon Musk’s xAI are also part of the program. But Anthropic’s A.I. chatbot, Claude, was the most widely used by the agency — and the only one on classified systems — thanks to its integration with technology from Palantir, a data analytics company that works with the federal government, according to defense officials with knowledge of the technology.
(The New York Times has sued OpenAI and Microsoft, accusing them of copyright infringement of news content related to A.I. systems. The companies have denied those claims.)
This year, Anthropic was set to formally sign a contract with the Department of Defense. That was when the company asked for limits to how its A.I. tools could be deployed.
Anthropic has long been more vocal than other A.I. companies on safety issues. In a podcast interview in 2023, Dr. Amodei said there was a 10 to 25 percent chance that A.I. could destroy humanity. Internally, the company has strict guidelines that bar its technology from being used to facilitate violence.
In January, Dr. Amodei wrote in an essay on his personal website that “using A.I. for domestic mass surveillance and mass propaganda” seemed “entirely illegitimate” to him. He added that A.I.-automated weapons could greatly increase the risks “of democratic governments turning them against their own people to seize power.”
In contract negotiations, the Defense Department pushed back against Anthropic, saying it would use A.I. in accordance with the law, according to people with knowledge of the conversations.
On Sunday, The Wall Street Journal reported that Anthropic’s technology had been used in January in the U.S. military operation to capture Venezuela’s president, Nicolás Maduro, and that Anthropic employees had raised concerns with Palantir about the role its technology played. Mr. Hegseth and others in the Pentagon were angered by what they saw as Anthropic’s resistance, current and former officials briefed on the matter said.
A Pentagon official said a “senior executive” from Anthropic had asked a senior official from Palantir if Anthropic’s A.I. model had been used in the Maduro raid. The Pentagon official said the exchange had alarmed Palantir because it implied that Anthropic might disapprove of its model’s being used in the operation. As a result, the Palantir executive contacted the Pentagon, the Pentagon official said.
Two people close to Anthropic said only one employee had raised questions about the Maduro operation with a Palantir employee in a routine meeting. Anthropic said it had “not discussed this with, or expressed concerns to, any industry partners outside of routine discussions on strictly technical matters.”
Palantir did not respond to a request for comment.
If Anthropic is categorized as a supply chain risk by the Defense Department, it will find itself in rare company. The designation is typically used only on firms that do business with China or take supplies from China.
Given that Claude and other Anthropic products are widely integrated across the department, it would be difficult to replace the technology, two officers who have used the technology said. Google’s and OpenAI’s A.I. contracts with the Pentagon are for unclassified systems.
On Feb. 9, OpenAI said in a blog post that it was expanding its work with the Pentagon. “It is important for the United States and other democratic countries to understand how, with the proper safeguards, AI can help protect people, deter adversaries, and prevent future conflict,” OpenAI wrote.
Google and xAI did not respond to requests for comment.
Despite the public fighting, negotiations between the Defense Department and Anthropic were continuing this week, two defense officials said.
Tyler Pager contributed reporting from Washington.
Sheera Frenkel is a reporter based in the San Francisco Bay Area, covering the ways technology affects everyday lives with a focus on social media companies, including Facebook, Instagram, Twitter, TikTok, YouTube, Telegram and WhatsApp.
The post Defense Dept. and Anthropic Square Off in Dispute Over A.I. Safety appeared first on New York Times.
