President Trump on Friday ordered all federal agencies to “immediately cease all use” of artificial intelligence startup Anthropic’s technology, significantly escalating the ongoing feud between the company and the Department of Defense.
In recent days, Anthropic, best known as the maker of ChatGPT’s leading competitor Claude, has been at loggerheads with the Department of Defense over the risks of deploying its powerful AI models for two controversial military purposes: developing fully autonomous weapons and conducting mass surveillance of American citizens.
“America will never allow radical left-wing corporations to determine how our great military fights and wins wars!” Trump wrote on social media. “The left-wing lunatics at Anthropic made a disastrous mistake in trying to fortify the War Department and force it to follow the Code of Service rather than the Constitution.”
But the battle between Anthropic and the Pentagon is about more than just a $200 million defense contract, and more than just drones and surveillance. It’s also a battle over the future of “the most revolutionary technology since the splitting of the atom” and a battle over who gets to control it, especially when it comes to killing or spying on humans.
In a closed-door meeting on Tuesday, Defense Secretary Pete Hegseth issued a blunt ultimatum to Anthropic CEO Dario Amodei: give him free access to your AI models for “all lawful uses” by 5:01pm ET on Friday, or face serious consequences.
Amodei has talked about safety more than any other AI giant and has resisted such demands since January. He insisted that Antropic wants to work with the Department of Defense. In fact, the Claude is currently the only model used in the department’s classified systems, an arrangement that dates back to 2024.
But Amodei said his company has long considered large-scale domestic surveillance a red line in ethics, and Claude isn’t ready to reliably and responsibly control fully autonomous weapons without human safeguards (at least not yet).
On Thursday, Amodei formally rejected Hegseth’s “best and last proposal,” writing that while he “deeply believes in the existential importance of leveraging AI to protect the United States and other democracies and defeat authoritarian adversaries,” he also believes that “in limited cases…AI could undermine rather than protect democratic values.”
The Department of Defense strongly opposed this. “It’s unfortunate that @DarioAmodei is a liar and has a God complex,” Emil Michael, a senior Pentagon official who oversees artificial intelligence, wrote in X late Thursday. “He wants nothing more than to try to personally control the U.S. military and is comfortable putting our nation’s security at risk. @DeptofWar will always abide by the law, but we will not bow down.” [the] It’s the whims of commercial technology companies. ”
Until Trump voiced his opinion, observers weren’t sure how Hegseth would react. On Tuesday, the defense secretary reportedly threatened two potential outcomes. The government could force Anthropic to hand over its technology by invoking the Defense Production Act, or it could block the AI giant from doing business with the Pentagon by blacklisting it and declaring it a “supply chain risk,” a penalty typically imposed on companies from adversaries such as China.
But then Trump instead moved to unilaterally ban anthropics from the entire federal government. “We don’t need it, we don’t want it, and we will never do business with them again!” he wrote on social media, adding that “there will be a six-month phase-out period for government agencies like the Department of the Army that use Anthropic’s products at various levels.”
The president also warned that there would be “significant civil and criminal consequences” if Anthropic “is not served during this phase-out period.” (Amodei previously vowed to “work to ensure a smooth transition to another provider.”)
Shortly after Friday’s 5:01 p.m. deadline passed, Hegseth made good on his Tuesday threat and directed the Pentagon to designate the human issue as a “supply chain risk to national security.”
“Effective immediately, any contractor, supplier, or partner doing business with the U.S. military may not engage in any commercial activity with Anthropic,” Hegseth wrote to X.
No other U.S. company has ever been declared a supply chain risk.
The Pentagon’s position is relatively simple. Under the Trump administration, the military has doubled its use of cutting-edge AI. In July, the Department of Defense awarded contracts worth up to $200 million each to four AI companies: Anthropic, OpenAI, Google DeepMind, and Elon Musk’s xAI. What are the department’s goals? A memo issued last month says it will transform the U.S. military into an “AI-first” force by quickly integrating top commercial AI models into combat, intelligence and support operations.
Historically, governments have always had the upper hand in these types of public-private partnerships. As Pentagon chief spokesman Sean Parnell said on Thursday’s X-Post, “We don’t let any company dictate the terms of how we make operational decisions.”
After all, Lockheed Martin hasn’t taught the Air Force how to fly the F-22, so Parnell’s assertion that the Pentagon is “not interested” in using AI to “conduct mass surveillance of Americans” or in “developing autonomous weapons that operate without human involvement” is enough, right?
“What we are asking is that the Department of Defense be allowed to use Anthropic models for any lawful purpose,” Parnell wrote. “This is a simple, common sense request to prevent Anthropic from potentially jeopardizing critical military operations and endangering our nation’s warfighters.”
However, experts say there are two reasons why AI is different from previous technologies.
First, we are progressing because of commerce, not because of government. “From nuclear propulsion to stealth to GPS, nations have been the primary drivers of discovery, and industry has been the integrator and manufacturer,” Rear Adm. Lorin Selby, former chief of naval research, recently told CNBC. “Today, the commercial sector is the primary driver of frontier capabilities…and the Department of the Army is no longer defining the limits of what is technically possible for artificial intelligence; it is adapting to it.”
This gives private companies like Anthropic more leverage than before.
Second, even the creators of AI, as Amodei once said, “don’t understand how the AI we create works or what it can do.” The risk, then, is not just that powerful AI will allow governments to “mockery” of Fourth Amendment privacy rights by collecting “scattered, individually innocuous data.” [about individual Americans] “We cannot rely on fully autonomous weapons to provide a comprehensive picture of human life automatically and at scale,” Amodei said, or, in matters of life and death, “to provide the critical judgment that highly trained, specialized forces make every day.”
And the risk is that whatever the term “all lawful purposes” encompasses today, it won’t be able to keep up with the capabilities of tomorrow’s AI.
“Previously we were asking for unconditional access.” [these] System readiness does not claim authority. “The gamble is that the unknown doesn’t matter,” said Thomas Wright, a senior fellow at the Brookings Institution. “The danger is not that Silicon Valley will wield too much power over the military; it is that neither side fully understands the systems they are rushing to implement, and the consequences of that ignorance will be tested out in the world, not in the laboratory.”
