If a person can refuse to fight in a war on moral grounds, what about a corporation?
It sounds absurd. But the legal scaffolding is already there.
Two landmark Supreme Court cases, Citizens United v. FEC (2010) and Burwell v. Hobby Lobby (2014), have extended constitutional rights to corporations that were once reserved for flesh-and-blood human beings. The first gave companies free speech. The second gave them sincerely held beliefs. Put those two precedents together, and you end up with an uncomfortable question that just became very relevant.
Whoa, wtf? Why is an AI blog talking about this? Because last week, AI company Anthropic refused the Pentagon's demand for unrestricted military use of its popular AI model, Claude. Anthropic drew two red lines: no mass surveillance of Americans, no fully autonomous weapons without human oversight. The Trump administration banned the company from federal use and designated it a "supply-chain risk to national security", a label normally reserved for companies like Huawei.
Anthropic has vowed to challenge the designation in court.
So here's the thought experiment: If the legal system has already decided that corporations can speak and corporations can believe, could a corporation also object?
P.S: Goes without saying, but this is not legal advice. We're not even saying we agree one way or the other on any of this. Just pontificating for the thought experiment.
The Legal Building Blocks
Let's start with what the Supreme Court has already told us corporations can do.
Citizens United (2010) held that corporations have First Amendment political speech rights. The majority opinion was explicit: "The First Amendment does not allow political speech restrictions based on a speaker's corporate identity." Corporations, as associations of individuals, can spend money to exercise those rights. Whether you love or hate the decision, it established a clear principle: a company can speak its mind.
Hobby Lobby (2014) went further. The Court ruled 5-4 that closely held for-profit corporations can exercise religion under the Religious Freedom Restoration Act (RFRA). The Greens, who own Hobby Lobby, argued that their company's health insurance plan shouldn't be required to cover certain contraceptives that violated their sincere religious beliefs. The Court agreed, holding that closely held corporations are legal "persons" capable of exercising religion, and that the government couldn't compel them to violate those beliefs without meeting the highest legal standard: a "compelling government interest" pursued through the "least restrictive means."
The key phrase is "sincerely held beliefs." The Court didn't require those beliefs to be scientifically correct or universally shared. They just had to be sincere. And the corporation had to be closely held, meaning a small group of individuals owned it and were deeply involved in its governance.
Sound like any AI companies you know?
The Conscientious Objector Framework
Conscientious objector status is rooted in the Selective Service framework. To qualify, an individual must demonstrate sincerely held religious or moral beliefs that oppose participation in war. Courts evaluate these claims through deeply personal inquiry: examining the person's conscience, the consistency of their beliefs, and their sincerity.
The framework has a few key features worth noting:
- It doesn't require religious belief. Since Welsh v. United States (1970), purely moral or ethical objections qualify, as long as they occupy a place in the objector's life "parallel to that filled by God" in traditionally religious people.
- It requires sincerity, not popularity. The government doesn't get to decide whether your beliefs are correct. It only gets to ask whether you actually hold them.
- The burden shifts. Once an individual establishes a sincere moral objection, the government must demonstrate a compelling need for that specific person's participation, and that there's no less restrictive way to meet that need.
Now, conscientious objector status has traditionally applied to individual humans being drafted into military service. No court has ever applied it to a company. That's the leap. But the precedents above keep shrinking the gap.
The Anthropic Case
If you were going to argue that any company in America qualifies as a "corporate conscientious objector," Anthropic is your best test case, and it's not particularly close:
- Consider the evidence of sincerity. Anthropic wasn't founded just to make competing chatbots to OpenAI.
- It was founded in 2021 by former OpenAI employees who left specifically over concerns about AI safety.
- The company's entire identity, from its name to its corporate structure to its "Responsible Scaling Policy," is built around the premise that powerful AI needs ethical boundaries.
- CEO Dario Amodei has staked his public reputation on the principle that "disagreeing with the government is the most American thing in the world."
This is not a company that discovered its moral convictions last Thursday. The documented history of those beliefs stretches back to founding. That matters. Courts evaluating conscientious objector claims look for exactly this: a consistent, long-standing pattern of belief, not a convenient last-minute pivot.
Now consider the closely held question. Anthropic is a private company. Its leadership, particularly the Amodei siblings, are deeply involved in governance. This mirrors the Hobby Lobby structure, where the Court ruled that closely held corporations can hold the sincere beliefs of their owners.
And consider the nature of the objection. Anthropic isn't refusing to work with the military entirely. It had a $200M Pentagon contract. Claude was the only AI model on the military's classified networks. The company agreed to every lawful military use except two specific applications: mass surveillance of Americans and fully autonomous weapons. That's not a blanket refusal to participate. That's a targeted moral objection to specific acts.
That specificity actually strengthens the case. The Selective Service framework recognizes both full objectors (who oppose all war) and selective objectors (who oppose specific conflicts or methods). Anthropic's position is analogous to someone willing to serve but refusing to participate in torture.
The Government's Counterargument
The strongest argument against corporate conscientious objector status is straightforward: the doctrine was designed for individual humans being conscripted into military service, not companies being asked to sell a product to the government. These are different things.
And that's fair. No one is putting Claude in uniform.
But the Defense Production Act (DPA) gives the government power to compel companies to produce goods for national defense. If Anthropic were ordered under the DPA to provide unrestricted access to Claude, the legal dynamics start looking more like compulsion than commerce. And the question becomes: if the government can compel a company to participate in something that violates its sincerely held beliefs, what protections does that company have?
After Hobby Lobby, the answer is: at least some.
The government would also need to clear the "least restrictive means" test. This is where the case gets interesting for Anthropic, because the government would need to show two things:
- It can't get the goods elsewhere. But it can. OpenAI signed its own Pentagon deal within hours of the Anthropic ban. Google's Gemini, xAI's Grok, and Meta's Llama are all available. The AI market is not a monopoly.
- This specific company's participation is essential. It is not. As much as Claude is impressive (and it is; we use it to write this newsletter), the Pentagon's operations do not depend on one company's chatbot. If you can't execute a military operation without a specific AI vendor, you have a planning problem, not a procurement problem.
The Biggest Weakness
Because the doctrine of conscientious objection was created to protect individual human conscience from state coercion, extending it to corporations risks diluting its moral weight.
Justice Ginsburg warned about exactly this in her Hobby Lobby dissent. She wrote that the majority's "expansive notion of corporate personhood" would invite for-profit entities to seek "religion-based exemptions from regulations they deem offensive to their faith." Expand that principle too far, and any company could claim moral grounds to dodge any law it finds inconvenient.
That's a real concern. The line between principled objection and strategic convenience is thinner than it looks.
But the counter to that concern is also real: if we've already decided that corporations can speak and that corporations can pray, we should at least be willing to ask whether they can also refuse to do things that violate what they claim to believe. The alternative is a version of corporate personhood that only runs in one direction, where companies get the rights of people when it benefits them but lose the obligations of conscience when it doesn't.
Where This Actually Goes
Nowhere, probably, at least for now. Anthropic is challenging its supply-chain risk designation in court, not filing for conscientious objector status. The legal theory we've outlined here is a thought experiment, not a brief.
But thought experiments do have a way of becoming briefs. Every major expansion of corporate rights in American law started as an idea that seemed absurd until it wasn't. "Corporations have free speech" was laughable until Citizens United. "Corporations can exercise religion" was bizarre until Hobby Lobby.
"Corporations can refuse to participate in war" might be next. Or it might remain forever in the category of interesting cocktail party arguments. Either way, the legal scaffolding is there, built brick by brick by courts that probably weren't thinking about AI when they laid each one. Man, so many of our laws need to get rewritten with an AI lens... it's crazy.
One thing is clear: the question Anthropic is forcing, whether a company can draw moral lines the government must respect, isn't going away. It's only going to get louder as AI gets more powerful, more embedded in military systems, and more capable of things that individual humans used to decide for themselves.
And if the past few weeks have taught us anything, it's that the US military agrees. Hours after banning Anthropic, it used Claude in combat anyway.
Not legal advice. The author has no legal training. But the author did ask Claude for help writing this, which feels relevant.
