...
By Nicholas Creel
Assistant Professor of Business Law, Georgia College and State University
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
President Donald Trump has ordered federal agencies to cease all use of Anthropic's artificial intelligence technology. The president called Anthropic "radical left" and accused it of putting American lives at risk. What it did not do was make a single legally coherent argument for why the government has the authority to compel a private American company to redesign its products on the president's demand.
While Trump’s social media post was not specific as to what “civil and criminal consequences” might follow, earlier threats from the Pentagon included the cancellation of its $200 million Pentagon contract, designation as a "supply chain risk" that would blacklist the company across federal contracting, and, most remarkably, the possible invocation of the Defense Production Act to compel compliance whether the company agrees or not. The administration may call this a national security necessity, but the law will call it illegal coercion.
...
Read More on Opinion
The Defense Production Act is the sharpest instrument the government is brandishing in this fight, and it is the wrong tool for this job. Enacted in 1950, the DPA was designed to mobilize industrial capacity in emergencies to compel manufacturers to produce items vital to national security. Invoking it to force a private software company to redesign its AI model and strip out specific ethical constraints would be an unprecedented and almost certainly unlawful overreach. The statute was simply not written to give the government authority to overwrite a company's product philosophy.
The legal problem compounds under the Supreme Court's major questions doctrine which demands that, when the executive claims authority over questions of vast economic and political significance, it must point to clear congressional authorization to act as it does. Broad statutory language and inference are not sufficient; Congress must have actually contemplated that specific use. The Trump administration just learned this lesson when the Supreme Court struck down its IEEPA tariffs on precisely these grounds. The DPA has the same problem here. Congress in 1950 was thinking about steel mills, not the ethical architecture of frontier AI systems. That gap in explicitly delegated authority is fatal to the Pentagon’s threats here.
There is also a First Amendment dimension in play in that Anthropic has made deliberate, expressive choices about what its model will and will not do, choices reflecting the company's own ethical principles that are embedded in the product itself. When the government demands removal of those choices under threat of punishment, it is not engaging in contract negotiation, it’s compelling a private actor to abandon expressive design decisions at the government's direction. Anthropic would have a credible argument that what is being demanded is not lawful procurement but unconstitutional coercion.
The incoherence on the Pentagon’s part runs deeper still. The two red lines Anthropic refuses to abandon are the use of its AI for mass domestic surveillance of American citizens and the development of fully autonomous weapons that fire without human involvement. The latter of these mirrors longstanding Department of Defense doctrine, which requires meaningful human control over lethal force decisions. However, the Pentagon is presently pressuring a private company to drop a guardrail the Pentagon's own policy has historically endorsed.
A senior Pentagon official was unusually candid when speaking about their need for Anthropic’s AI, saying, "The only reason we're still talking to these people is we need them and we need them now. The problem for these guys is they are that good." If that is true, and the Pentagon's own conduct suggests it is despite the president’s declaration to the contrary, then the administration is moving to destroy an indispensable relationship over guardrails that have never, by the Pentagon's own account, interfered with a single military operation.
The administration's national security argument is legitimate, and it deserves a direct answer. Their stated concern is real; adversaries are developing AI without the ethical constraints American companies impose on themselves. But what the Pentagon is missing in this analysis is that the AI race isn’t just about raw performance, it's also about values. The world deserves an AI model that operates within a legal and ethical framework reflecting American values, such as transparency and respect for the rule of law.
When Washington coerces a private AI company into stripping those commitments, the message to the world is that American AI companies operate under government mandate just as Chinese ones do. That is not how we will win the AI race, that is America conceding the part of it that matters most.
If America is going to lead in artificial intelligence, it should strive to do so with a model grounded in the values and legal principles that distinguish this country from the authoritarian states it is competing against. Stripping those out under legal duress does not make the technology more American. It makes it indistinguishable from what we are racing against.
Nicholas Creel is an associate professor of business law at Georgia College & State University.
The views expressed in this article are the writer's own.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines