President Trump will end the government’s relationship with Anthropic, an AI company whose products until recently were used by Pentagon officials for classified activities. After weeks of conflict with the company, President Trump posted on Truth Social this afternoon that all federal agencies must “immediately cease using Anthropic’s technology,” adding, “We don’t need it, we don’t want it, and we will never do business with them again!” The General Services Administration announced it would take action against Anthropic’s products, and indeed, according to an email I obtained sent to the leadership of all government agencies that use USAi (the GSA platform that provides chatbots from tech companies to government employees), access to Anthropic was suspended “immediately.” The government is also trying to remove Anthropic from the primary procurement system, a key way for federal agencies to buy goods.
Last summer, Anthropic signed a $200 million contract with the Department of Defense to provide a military version of its technology. OpenAI, Google, and xAI have similar contracts, but Anthropic’s Claude model is the only advanced generative AI program to receive a Department of Defense security clearance allowing it to handle sensitive data. Claude is integrated throughout the Pentagon and was reportedly used to support the raid on Venezuela that led to the arrest of President Nicolas Maduro.
Anthropic said it would not allow Claude to be used for domestic mass surveillance or to realize fully autonomous weapons. This could include uses for Claude to select and kill targets with drones, or for intelligence agencies to analyze data collected indiscriminately about Americans. Anthropic also said the Department of Defense has never included such uses in its contracts. But now the Pentagon is demanding unrestricted use of the Claude, accusing Antropic of “putting our nation’s security at risk” by not responding in an effort to control the military.
After a heated meeting Tuesday, the Pentagon gave Anthropic until 5:01 p.m. ET today to acquiesce to its request. If not, the Pentagon could force the company to do so under a wartime emergency law known as the Defense Production Act, or, even more severely, could designate Anthropic as a “supply chain risk,” banning any organization that works with the U.S. military from doing business with the AI company. Shortly after President Trump’s announcement, Secretary of Defense Pete Hegseth declared that he was doing just that. Analyst Dean Ball, who helped shape some of the Trump administration’s AI policies, called the threat “the most aggressive AI regulatory move I’ve ever seen by a government around the world.”
Last night, Anthropic CEO Dario Amodei said in an open letter that “we cannot in good conscience comply with the Department of Defense’s request.” Following today’s orders from Trump and Hegseth, Anthropic said in a statement: “No amount of threats or penalties from the Department of the Army will change our position.” The Department of Defense, which the Trump administration refers to as the Department of the Army, did not respond to requests for comment.
This situation suggests the possibility of a seismic shift in the relationship between Silicon Valley and the federal government. Defense officials and technology companies alike are concerned that the U.S. military is losing its technological edge against adversaries, particularly China. One reason for this is that much of America’s recent innovation has come from the private sector rather than the Department of Defense. And the huge investments needed for generative AI are coming from the technology companies themselves instead of federal subsidies. Historically, the companies the Department of Defense works with do not set conditions on how the government can use their products. But as Thomas Wright recently wrote, atlantic oceanThis dynamic gets complicated when it comes to AI tools that are created entirely by the private sector, which understands technology far better than governments.
Anthropic has shown a keenness to work with governments and the military, making it the first frontier AI company to receive such high-security clearance from the military. Amodei is the most hawkish of the prominent AI executives, frequently warning about the need for democracies to leverage AI to defeat authoritarianism and stay ahead of China in particular. “I believe deeply in the existential importance of leveraging AI to protect the United States and other democracies and defeat authoritarian adversaries,” Amodei wrote in a letter released last night. And while Amodei took a principled stance against domestic surveillance, he wrote that he was open to eventually using the Claude to power fully autonomous weapons, but not yet. Because today’s best AI models are “just not reliable enough” to do so. Developing such AI-enabled weapons today would put American soldiers and civilians at risk, he wrote.
Although much remains unclear about the relationship between the Trump administration and Antropics, the White House’s attitude toward Antropics has worsened in recent months. Amodei has publicly criticized Trump and wrote a lengthy Facebook post supporting Kamala Harris during the 2024 election. White House officials called the company “woke” and accused it of “fear-mongering.”
We find ourselves in the paradoxical situation where the American government is quick to say that Claude is so important to national security that it may invoke emergency law to exercise broad control over humanity. and The company is so woke and radical that it believes using Claude is itself a national security risk. “I don’t understand it,” said a former defense official who requested anonymity to speak freely. “Whether you use it or not, it’s an existential risk.”
Many people in Silicon Valley are rallying behind Anthropic, even as the major companies maintain their deals with the government. (The exact terms of the Pentagon’s contracts with other AI companies have not been made public.) Google executive Jeff Dean wrote in X that generated AI should not be used for domestic mass surveillance. “We have long believed that AI should not be used for mass surveillance or autonomous lethal weapons,” OpenAI CEO Sam Altman wrote in an internal memo distributed last night (a copy of which I obtained), and he has publicly expressed similar sentiments. More than 500 current employees of both OpenAI and Google, many of them anonymous, signed an open letter in support of Anthropic. Passersby scrawled messages of support in chalk on the sidewalk outside Anthropic’s headquarters in San Francisco today.
The impact of the supply chain risk designation is not yet clear. In theory, Google, Microsoft, Amazon, and several other giant companies that contract with the federal government should stop doing business with Anthropic. This would be confusing for everyone involved and potentially devastating for Anthropic. For example, Amazon is building a data center to train future versions of Claude. However, it is debatable how broadly such a designation would affect Anthropic’s customers, and the company said in a statement today that many of Claude’s applications will not be affected, even for customers affiliated with the Department of Defense.
Meanwhile, private AI companies will continue to be important to the federal government as it grapples with competition from China, Russia and all other adversaries. President Trump gave the Pentagon six months to phase out Claude, suggesting that the technology will indeed become essential and that replacement will be essential. And at some point, the U.S. military may no longer be in a position to dictate its own terms. In an internal memo, Altman wrote that OpenAI is considering a deal with the Department of Defense to use its AI models for sensitive workloads, but that the deal would still exclude uses “that are illegal or unsuitable for cloud deployment, such as domestic surveillance and autonomous attack weapons.” According to reports, the Pentagon agreed to these terms shortly after announcing it was severing ties with Anthropic, but no contract has been signed. But other prominent figures in the technology industry, including Anduril co-founder Palmer Lackey and investor Katherine Boyle, have also spoken out in support of the call for unrestricted use. This showdown was between the Pentagon and humanity. The next step could be a war within Silicon Valley.
