The Pentagon vs. Anthropic: When "Red Lines" Become a National Security Question
The dispute between the U.S. Department of Defense and Anthropic has escalated into something that deserves careful disentangling. What looks like a straightforward contract disagreement is actually three different arguments happening simultaneously – and until those arguments are separated, the debate will continue generating more heat than light.
What Actually Happened
The facts, as reported, are these: Anthropic signed a $200 million contract with the Pentagon last summer to deploy its AI technology within classified systems. During subsequent negotiations, Anthropic stated it did not want its AI systems used for mass surveillance of Americans, and that the technology wasn't ready for use in targeting or firing decisions of lethal weapons.
The Pentagon's response was not to end the contract. Instead, Defense Secretary Pete Hegseth designated Anthropic a "supply-chain risk" – a label typically reserved for foreign adversaries. This designation requires any company or agency working with the Pentagon to certify it doesn't use Anthropic's models.
Anthropic sued. The DOD filed a 40-page response arguing that Anthropic poses an "unacceptable risk to national security" because the company might "attempt to disable its technology or preemptively alter the behavior of its model" during "warfighting operations" if its corporate "red lines" are crossed.
A hearing on Anthropic's request for a preliminary injunction is set for next Tuesday.
Three Arguments Masquerading as One
This dispute contains at least three distinct disagreements that participants keep conflating:
The contractual question: Can a vendor negotiate terms of use, or must contractors accept "all lawful purposes" language? This is a question about procurement norms and the boundaries of government contracting.
The authority question: Who decides how military AI is used – the company that built it, or the military that deploys it? This is a question about the appropriate relationship between private technology providers and state power.
The retaliation question: Did the DOD use an extraordinary legal mechanism (supply-chain-risk designation) to punish a company for its negotiating position? This is a question about whether the government acted in good faith or weaponized regulatory authority.
These are different questions with different answers. Someone could believe vendors should accept broad use terms (answering the contractual question one way) while also believing the supply-chain-risk designation was an inappropriate response (answering the retaliation question another way). The debate becomes incoherent when these positions are bundled together.
The Strongest Version of Each Position
The DOD's argument, in its strongest form, runs like this: Military operations require absolute reliability. A technology provider that reserves the right to define "red lines" introduces uncertainty into warfighting systems. The military cannot afford to wonder whether its AI tools will function during a crisis because a private company has ethical objections to a particular operation. This isn't about punishing Anthropic – it's about ensuring operational integrity.
Anthropic's argument, in its strongest form, runs like this: The company never claimed the right to disable systems during operations. It sought to negotiate terms before deployment, which is normal vendor behavior. The DOD's response – treating an American AI company like a foreign adversary – is wildly disproportionate to a contract negotiation. As CEO Dario Amodei stated in late February:
"Anthropic understands that the Department of War, not private companies, makes military decisions. We have never raised objections to particular military operations nor attempted to limit use of our technology in an ad hoc manner."
Constitutional lawyer Chris Mattei, a former Justice Department attorney, argues the DOD's position collapses under scrutiny:
"The government is relying completely on conjectural, speculative imaginings to justify a very, very serious legal step they've taken against Anthropic."
He notes there has been no investigation to support the DOD's concerns about Anthropic potentially disabling or altering its AI models during operations.
Where Each Position Breaks Down
The DOD's position has a significant weakness: if the concern is operational reliability, the appropriate response is to end the contract and find another vendor – not to designate the company a supply-chain risk. The supply-chain-risk mechanism exists to protect against adversarial actors, not to resolve commercial disputes. Using it this way raises the question Mattei poses: why didn't the DOD simply "articulate a credible or even comprehensible rationale for why Anthropic's refusal to agree to an 'all lawful use' provision rendered it a supply chain risk as opposed to a vendor that DOD simply didn't want to do business with"?
Anthropic's position also has complications. The company's "red lines" language – however reasonable in a commercial context – does create genuine uncertainty about what happens if the military's use of AI evolves in ways Anthropic finds objectionable. The DOD's concern isn't entirely fabricated, even if its response is disproportionate.
The Broader Stakes
This dispute matters beyond the immediate parties for several reasons.
First, it establishes precedent for how AI companies can negotiate with government clients. If Anthropic loses, the message to the industry is clear: accept "all lawful purposes" language or face existential regulatory consequences. OpenAI has already signed a deal allowing military use of its systems for "all lawful purposes" – a phrase some employees have noted could encompass exactly the uses Anthropic sought to avoid.
Second, it tests whether the supply-chain-risk designation can be used as a general-purpose enforcement tool against domestic companies. Dean Ball, a former Trump White House AI adviser, has called the designation a "death rattle" of the American republic, arguing the government has abandoned strategic clarity in favor of "thuggish" tribalism that treats domestic innovators worse than foreign adversaries.
Third, it reveals the absence of clear frameworks for AI governance in military contexts. The dispute exists partly because there are no established norms for what AI companies can and cannot negotiate when contracting with defense agencies.
The Question Worth Asking
The debate has largely focused on whether Anthropic's position is reasonable or whether the DOD's response is proportionate. But perhaps the more productive question is: what institutional arrangements would prevent this kind of dispute from arising in the first place?
Several tech companies and employees – including from OpenAI, Google, and Microsoft – have filed amicus briefs supporting Anthropic. Hundreds of employees from OpenAI and Google have urged the DOD to withdraw its designation and called on Congress to push back. This suggests the industry recognizes that the precedent being set affects everyone.
The irony is that Anthropic was, until this dispute, the only frontier AI lab with classified-ready systems. The U.S. military was relying on Claude in its Iran campaign, where American forces use AI tools to manage operational data. The supply-chain-risk designation disrupts not just Anthropic but the Pentagon's own operations.
This is what happens when a disagreement about terms becomes a disagreement about loyalty, and a disagreement about loyalty becomes a question of national security. The categories keep escalating, but the underlying issue – how should AI companies and governments negotiate the boundaries of military AI use? – remains unresolved.
That question won't be answered in a California federal court. It requires the kind of sustained, multi-stakeholder conversation that neither side of this dispute has been willing to have. The hearing next Tuesday will determine Anthropic's immediate fate. The longer-term question of how democracies govern military AI remains wide open.
These are precisely the governance questions that demand more than headlines and hot takes. For those ready to engage with the institutional frameworks Europe is building – and the lessons it might offer – Human x AI Europe convenes in Vienna on May 19. The conversation continues there.
Frequently Asked Questions
Q: What is a supply-chain-risk designation?
A: A supply-chain-risk designation is a legal mechanism the U.S. Department of Defense uses to identify entities that pose security threats. It requires any company or agency working with the Pentagon to certify it doesn't use products from the designated entity. This designation is typically reserved for foreign adversaries.
Q: What specific "red lines" did Anthropic establish with the Pentagon?
A: Anthropic stated it did not want its AI systems used for mass surveillance of Americans and that its technology wasn't ready for use in targeting or firing decisions of lethal weapons. The company sought to negotiate these terms before deployment, not to intervene during operations.
Q: When is the court hearing on Anthropic's preliminary injunction request?
A: The hearing is scheduled for Tuesday, March 24, 2026, in a California federal court. Anthropic is seeking to temporarily block the DOD from enforcing its supply-chain-risk designation.
Q: How does this affect the Pentagon's current military operations?
A: The designation disrupts Pentagon operations because Anthropic's Claude was being used in the U.S. military's Iran campaign and was installed in Palantir's Maven Smart System, which military operators in the Middle East rely on. Anthropic was the only frontier AI lab with classified-ready systems.
Q: What is OpenAI's position on military AI use compared to Anthropic's?
A: OpenAI signed a deal with the Department of Defense allowing military use of its AI systems for "all lawful purposes" – broader language than Anthropic was willing to accept. Some OpenAI employees have expressed concern that this phrasing could permit the uses Anthropic sought to avoid.
Q: What legal argument is Anthropic making in its lawsuit?
A: Anthropic argues the DOD infringed on its First Amendment rights and punished the company based on ideological grounds. Constitutional lawyer Chris Mattei argues the government is relying on "conjectural, speculative imaginings" without evidence that Anthropic would actually disable or alter its AI models during military operations.