People throw around the phrase “corporations are people” so often that it has started to sound almost meaningless. But my concern is not the slogan itself. My concern is what follows from it when power begins to concentrate in the executive branch.
If the government can treat a corporation as a rights-bearing legal person when that is useful, can it also treat that corporation as something to be disciplined, blacklisted, or crushed when it refuses to comply? And if it can do that, what happens to the rights and autonomy of the actual human beings whose speech, labor, conscience, and livelihoods are bound up with that institution? That is the real question raised by the legal fight between Anthropic and the Department of War. [1]
This is not merely a contract dispute. It is a test of whether the executive branch can use the language of national security to punish a private company for drawing ethical lines around the use of its technology. In a March 26 preliminary injunction, Judge Rita Lin wrote that the government likely had “no legitimate basis” to brand Anthropic a supply-chain threat, called the theory “Orwellian,” said Anthropic likely suffered a due process violation because it had no notice or opportunity to respond, and found likely First Amendment retaliation. That is not routine judicial language. It is the language of a court staring directly at a probable abuse of power. [1]
To be clear, corporate personhood is a legal fiction, not a declaration that companies are human beings. Corporations can own property, sign contracts, sue and be sued, and in some contexts assert constitutional protections. But they do not vote, raise families, or possess the full bundle of human rights. Even Citizens United, for all the controversy around it, did not say corporations and citizens are the same thing. It held that the government cannot restrict certain political expenditures simply because the speaker uses the corporate form. [4]
That distinction matters because the danger is not that corporations have somehow become too human. The danger is that the state can use power over corporations to gain leverage over the people behind them.
In modern life, people do not exercise rights only as isolated individuals. They exercise them through institutions. Through employers, publishers, labs, universities, nonprofits, contractors, and companies. Control the institution, and you do not need to go after every individual one by one. You can chill them instead. You can make examples of them. You can teach everyone else what resistance will cost.
That is why the allegations in the Anthropic case matter so much. Judge Lin made clear that the government is free to stop using Anthropic’s products and find another vendor willing to accept “all lawful uses.” But the court found that the problem likely began when the government went further — when it allegedly used blacklist-style measures and a “supply chain risk” designation not because Anthropic was compromised or sabotaging systems, but because it publicly refused certain uses of its models, specifically mass domestic surveillance and fully autonomous lethal warfare. [1]
That is a very different thing from ordinary procurement discretion. A customer can decide not to buy. The state can choose another vendor. But once the executive starts using national-security labeling and broad coercive pressure to punish a company for saying no, the issue is no longer purchasing. It is power.
And American law is not silent on that point. In Youngstown Sheet & Tube Co. v. Sawyer, the Supreme Court rejected President Truman’s attempt to seize steel mills during wartime without proper legal authority. The principle was simple and enduring: invoking national defense does not magically create powers the Constitution and Congress did not grant. In NRA v. Vullo, the Court reaffirmed that government officials cannot coerce private parties into cutting ties with a disfavored speaker in order to punish or suppress protected expression. Those cases are not identical to Anthropic’s. But they point in the same direction. The executive does not get a free pass to route around constitutional limits simply by acting through pressure rather than direct command. [2][3]
This is where the real constitutional concern comes into view. Once the state can punish a corporation not for fraud, espionage, or breach, but for refusing a politically charged demand, then corporate rights are no longer the central issue. The real issue is the spillover. What happens to the engineers who object? What happens to executives who resist? What happens to employees, investors, vendors, researchers, and customers who learn that some lines are simply too expensive to hold?
Their rights may remain intact on paper. Their practical ability to exercise them begins to shrink.
That is one of the quieter ways free societies can be weakened. Not always by jailing dissidents in the first instance, but by gaining leverage over the institutional chokepoints through which dissent, work, publication, and association occur. Pressure the company rather than the worker. The publisher rather than the writer. The platform rather than the speaker. The contractor rather than the engineer. It is cleaner. More deniable. And in a modern administrative state, often more effective.
AI makes this more dangerous because the stakes are not trivial. These systems are not just office software. They are tools with obvious military, surveillance, intelligence, and policing applications. That means the question is bigger than Anthropic, or one administration, or one contract. It is about whether the government can force AI developers to abandon meaningful red lines and then punish them when they refuse. If that becomes normal, then we are not merely talking about technology policy anymore. We are talking about the architecture of state power. Civil-liberties groups have warned that compelled deployment of AI in surveillance contexts could pose serious threats to privacy and democratic freedom. They are right to worry. [5]
I am not saying the United States has already crossed into authoritarianism. That kind of language can be satisfying, but it can also be sloppy. What I am saying is more precise, and in some ways more alarming: this case appears to present an attempt to normalize an authoritarian method.
The method is simple. Recast disagreement as risk. Recast refusal as sabotage. Invoke national security. Bypass the ordinary safeguards. Punish broadly enough that everyone else in the sector understands the lesson.
That is not yet dictatorship. But it is the kind of mechanism by which democratic limits are softened, then bypassed, then forgotten.
The deepest issue here is not whether corporations are people. It is whether the government can use the corporate form as a back door to control people. If the answer ever becomes yes, then many of the rights we congratulate ourselves for having will remain in our books while fading in practice. The real test of a free society is not whether rights exist in theory. It is whether power is prevented from quietly routing around them. The Anthropic case matters because it may help decide exactly that. [1]
Notes
[1] Anthropic PBC v. U.S. Department of War et al., Order Granting Motion for Preliminary Injunction, U.S. District Court for the Northern District of California, March 26, 2026.
[2] Youngstown Sheet & Tube Co. v. Sawyer, 343 U.S. 579 (1952).
[3] National Rifle Association of America v. Vullo, 602 U.S. ___ (2024).
[4] Citizens United v. Federal Election Commission, 558 U.S. 310 (2010).
[5] American Civil Liberties Union and Center for Democracy & Technology, “ACLU and CDT Urge Court to Stop Government from Punishing Anthropic for Important Advocacy on AI Guardrails.”
Brennan Center for Justice, “The History of Corporate Personhood.”
O’Hare Truck Service, Inc. v. City of Northlake, 518 U.S. 712 (1996).