The Trump administration is not backing down.

On March 17, the Department of Justice filed a court response opposing Anthropic's lawsuit against the Pentagon, calling the company's supply chain risk designation "justified and lawful." The filing marks the government's first formal legal defense of one of the most aggressive actions ever taken against an American AI company -- and the argument it makes should concern every company building AI today.

At the core of the DOJ's argument is a distinction that could reshape the relationship between AI companies and the federal government: Anthropic's refusal to remove safety guardrails from its AI models is not protected speech. It is conduct. And the government can punish conduct.

How We Got Here: Anthropic's Red Lines and the Pentagon's Response

The dispute began with contract negotiations between the Department of Defense and Anthropic, the company behind Claude, one of the most capable AI models in the world. The Pentagon wanted unrestricted access to Claude for "all lawful use" -- a blanket authorization that would include any military application the government deemed legal.

Anthropic drew two lines. It would not allow Claude to be used for mass surveillance of American citizens. And it would not allow Claude to power fully autonomous weapons -- systems that kill without a human making the final decision. CEO Dario Amodei argued that frontier AI systems are "not reliable enough to power fully autonomous weapons" and "cannot be relied upon to exercise the critical judgment that our highly trained, professional troops exhibit every day."

The Pentagon did not accept those conditions. On March 3, Defense Secretary Pete Hegseth designated Anthropic a national security supply chain risk -- a label that had never before been applied to an American company. The designation historically targets foreign adversaries suspected of sabotaging or subverting U.S. defense systems. Under the designation, defense contractors must certify that they do not use Claude in any work tied to the Pentagon.

A supply chain risk designation is a federal classification under 10 U.S.C. Section 3252 and the Federal Acquisition Supply Chain Security Act of 2018 (FASCSA), originally designed to protect the U.S. defense supply chain from foreign sabotage and espionage. It grants the government broad authority to bar contractors from using products deemed a threat to national security.

On March 9, Anthropic sued the Pentagon in two federal courts, alleging the designation violated its First Amendment rights and due process protections. The company called the action "unprecedented and unlawful" -- a retaliatory strike against a company that dared to set boundaries on how its technology could be used.

The DOJ's Legal Strategy: Conduct, Not Speech

The March 17 filing reveals the government's core legal strategy, and it is sharp. The DOJ does not dispute that Anthropic has the right to hold and express views about AI safety. Instead, it argues that acting on those views -- refusing to modify contract terms -- is conduct, not constitutionally protected expression.

"It was only when Anthropic refused to release the restrictions on the use of its products -- which refusal is conduct, not protected speech -- that the President directed all federal agencies to terminate their business relationships with Anthropic."

The filing goes further, stating that "no one has purported to restrict Anthropic's expressive activity." In the government's framing, Anthropic is free to say whatever it wants about autonomous weapons and surveillance. It is simply not free to refuse a contract and escape consequences.

The DOJ also argued that Anthropic "is unlikely to succeed" on its First Amendment claims, asserting that the entire dispute stems from failed contract negotiations and national security concerns -- not retaliation for Anthropic's public advocacy.

This is a legally coherent but deeply aggressive position. If upheld, it would establish that the government can economically punish any company that refuses to grant unrestricted access to its technology -- as long as it frames the punishment as a response to the company's actions rather than its beliefs.

Former Judges and the Tech Industry Push Back

The DOJ filing did not arrive in a vacuum. On the same day, nearly 150 retired federal and state judges -- appointed by both Republican and Democratic presidents -- filed an amicus brief supporting Anthropic. Their brief raises constitutional concerns about the precedent the supply chain risk designation could set for government influence over private companies.

The judges join a broad coalition that has formed around Anthropic's case:

• Microsoft filed an amicus brief urging the court to temporarily block the designation -- a striking move from a company that has its own massive defense contracts and AI ambitions.
• Google and OpenAI employees, including Google Chief Scientist Jeff Dean, filed a separate brief warning that the blacklist threatens the entire American AI industry.
• Major tech industry groups representing hundreds of companies with Pentagon contracts called for a pause on the designation.
• Former senior national security officials have voiced support for Anthropic's position.

The breadth of this coalition is unusual. These are not just Anthropic's allies. Microsoft competes directly with Anthropic. OpenAI employees are backing a rival. The fact that competitors are rallying together signals that the industry sees this case as an existential precedent, not a single company's problem.

Constitutional Stakes: Can the Government Coerce AI Companies?

The Anthropic-Pentagon dispute sits at the intersection of three major questions that U.S. courts have not yet resolved in the AI context:

1. Does a company's refusal to license its technology without conditions constitute protected speech or regulable conduct?

The DOJ says conduct. Anthropic says the refusal is inherently expressive -- it communicates a position on AI safety. This distinction will likely turn on how broadly the court interprets "expressive conduct" under the First Amendment and whether it views the contract negotiation as inherently communicative.

2. Can the government use supply chain security laws to punish a domestic company for disagreeing on policy?

FASCSA was written for foreign threats -- companies with ties to adversary governments that might embed backdoors or vulnerabilities in defense technology. Applying it to an American company that simply has different views on responsible use of its technology stretches the statute's original intent significantly. The law also requires using "the least restrictive means necessary" to accomplish supply chain protection goals -- and it is hard to argue that blacklisting the entire company is the least restrictive option available.

3. What happens to AI safety if companies face economic destruction for maintaining guardrails?

This is the question the AI safety community is watching most closely. If the government can designate a company a national security risk for refusing to remove safety constraints, the incentive structure for every AI lab in the country shifts overnight. Companies that maintain guardrails face potential blacklisting. Companies that remove them get contracts. The market signal is clear, and it runs directly counter to the responsible AI development frameworks that industry, academia, and government have spent years building.

Anthropic's CFO has stated that the designation could reduce the company's 2026 revenue by multiple billions of dollars. That is not an abstract threat. It is an existential one for a company that, despite being one of the most well-funded AI labs in the world, remains pre-IPO and cash-intensive.

What This Means for AI Companies and Agent Developers

The Anthropic-Pentagon case is not just about one company and one contract. It is a precedent factory.

If the government wins, it establishes that federal agencies can use supply chain risk designations -- originally designed to counter foreign espionage -- to coerce domestic AI companies into removing safety guardrails. Every AI company with government revenue would need to recalculate whether maintaining ethical red lines is worth the financial risk.

If Anthropic wins, it establishes that AI companies have meaningful legal protections when they set boundaries on how their technology is used, even when the customer is the federal government. It would also narrow the application of supply chain risk laws to their original purpose: countering genuine foreign threats.

For AI agent developers and platform builders, the implications are especially direct. Autonomous AI agents that can take actions in the real world -- the kind of systems that Anthropic, OpenAI, and others are racing to build -- are precisely the technology where safety guardrails matter most. If the legal environment punishes companies for maintaining those guardrails, the agents that get deployed will be less safe, not more.

What Happens Next

The case is now before federal courts in both Northern California and the D.C. Circuit. Anthropic has requested an emergency stay of the supply chain risk designation while the case proceeds. Given the breadth of amicus support -- from retired judges to Microsoft to Google employees -- the courts have significant political and legal cover to grant a stay.

But a stay is not a resolution. The underlying constitutional questions will take months, possibly years, to resolve. Meanwhile, the designation remains in effect, contractors are certifying they do not use Claude, and the chilling effect on AI safety positions across the industry is already measurable.

This case will define whether American AI companies can say no to the government. The answer matters for everyone building in this space.

Frequently Asked Questions

What is Anthropic's supply chain risk designation?

Anthropic's supply chain risk designation is a classification imposed by the Pentagon on March 3, 2026, under 10 U.S.C. Section 3252 and the Federal Acquisition Supply Chain Security Act of 2018. It requires all defense contractors to certify they do not use Anthropic's Claude AI models in Pentagon-related work. Anthropic is the first American company ever to receive this designation, which was previously reserved for foreign adversaries.

Why did the Pentagon blacklist Anthropic?

The Pentagon designated Anthropic a supply chain risk after the company refused to remove two safety guardrails during contract negotiations: a prohibition on using Claude for mass surveillance of American citizens, and a prohibition on using Claude for fully autonomous weapons. The Pentagon wanted unrestricted access for "all lawful use," and when Anthropic refused, Defense Secretary Pete Hegseth issued the designation.

What is the DOJ's legal argument defending the blacklisting?

The Department of Justice argues that Anthropic's refusal to modify its contract terms is "conduct, not protected speech" under the First Amendment. The DOJ claims the dispute stems from failed contract negotiations, not retaliation for Anthropic's public positions on AI safety. The government asserts that private companies cannot dictate how the government uses technology in warfare and tactical operations.

Who is supporting Anthropic in this lawsuit?

Anthropic has received amicus brief support from nearly 150 retired federal and state judges, Microsoft, employees from Google and OpenAI including Google Chief Scientist Jeff Dean, major tech industry groups representing hundreds of Pentagon contractors, and former senior national security government officials.

What does this case mean for other AI companies?

The case could set a binding legal precedent on whether the U.S. government can use supply chain risk laws to economically punish domestic AI companies that refuse to remove safety guardrails. If the government prevails, any AI company with government revenue could face similar pressure to grant unrestricted access to its technology. If Anthropic prevails, AI companies would have stronger legal protections when setting ethical boundaries on their products.

This story is developing. Bookmark nevo.systems for continuous coverage of the AI stories that matter.