Sciensify
On February 27, 2026, at 5:01 p.m. Eastern, a deadline passed that the U.S. Department of Defense had set for Anthropic: agree to the Pentagon's terms for using Claude, or face consequences.
Anthropic's answer was no.
Within hours, President Trump directed all federal agencies to cease using Anthropic's AI technology. Defense Secretary Pete Hegseth designated Anthropic a "supply chain risk" — a label previously reserved for companies suspected of being instruments of foreign governments.
Then OpenAI CEO Sam Altman announced, in what is one of the more remarkable pieces of timing in recent tech history, that his company had just secured the Pentagon contract Anthropic had walked away from.
This is the story of what led there, what it means for AI policy, and what it means for anyone building on AI tools.
How It Started: A $200 Million Deal and a Handshake Agreement
The relationship between Anthropic and the U.S. government was not adversarial at the outset.
In July 2025, Anthropic and the Pentagon signed a contract valued at up to $200 million. According to AI Thinker Lab's full timeline, Claude became the first frontier AI model approved for use on classified U.S. military networks. The contract included a key stipulation: the Department of Defense agreed to abide by Anthropic's Acceptable Use Policy.
That policy contained two explicit restrictions — the ones that would eventually blow the deal apart.
Claude was not to be used to power fully autonomous weapons systems. And Claude was not to be used in the mass surveillance of U.S. citizens.
Anthropic has since referred to these as its "redlines" — the two categories of use it will not enable regardless of who the customer is or how much money is on the table.
For several months, the arrangement held. Then the Pentagon came back with a renegotiation request.
The Demand: "All Lawful Use"
The Department of Defense, now operating under a more aggressive posture toward AI autonomy, asked Anthropic to waive the two restrictions.
The new term the Pentagon proposed was "all lawful use" — meaning the government would determine what was and was not acceptable, without Anthropic's guidelines as a limiting condition. CNN reported that the DOD wanted "unfettered access to Claude across all lawful purposes."
Anthropic declined. The two sides entered weeks of negotiations that produced no agreement.
The Pentagon set a hard deadline: 5:01 p.m. on February 27, 2026. Anthropic's answer before that deadline, in the company's own words published on its website: "Threats do not change our position: we cannot in good conscience accede to their request."
The Designation: Unprecedented and Retaliatory
What followed was, by legal analysis, without precedent.
Mayer Brown's legal assessment described the supply chain risk designation as the first time the United States had ever applied the label to an American company, and the first time it had been used as apparent retaliation for a business failing to agree to government contracting terms.
The designation is significant beyond symbolism. Defense contractors working with the Pentagon are typically prohibited from using suppliers or vendors designated as supply chain risks. By labeling Anthropic this way, the government effectively required every defense contractor building on Claude to find a different AI model — giving them six months to transition away from Anthropic's products entirely.
The government's own records, obtained during subsequent litigation, stated that Anthropic was designated a supply chain risk because of its "hostile manner through the press." The company had spoken publicly about its refusal to comply with the Pentagon's demands before the deadline passed.
That detail would later become legally significant.
The Court Fight: A First Amendment Win — Then a Setback
Anthropic did not accept the designation quietly. On March 9, 2026, the company filed two federal lawsuits challenging the action. The Washington Post confirmed the filings.
On March 26, CNBC reported that Judge Rita F. Lin issued a preliminary injunction in Anthropic's favor. Her ruling held that designating Anthropic as a supply chain risk because of its public communications about the dispute constituted textbook First Amendment retaliation. "Punishing Anthropic for bringing public scrutiny to the government's contracting position is classic illegal First Amendment retaliation," the ruling stated.
The injunction was a meaningful win — but temporary.
On April 8, CNBC reported that an appeals court denied Anthropic's bid to maintain the block while the full case proceeds. The designation is currently active.
As of May 19, CNBC confirmed that opening arguments in the main case began in a Washington D.C. federal court. The legal fight is ongoing.
The Industry Response: Rare Solidarity
The designation triggered an unusual response from an industry that rarely breaks ranks with government power.
More than 30 researchers and engineers from Google DeepMind and OpenAI — including Google's chief scientist Jeff Dean — filed an amicus brief supporting Anthropic's position in the litigation. Fortune reported that the brief represented a direct challenge by employees of Anthropic's direct competitors to the government's action against the company.
At OpenAI, the situation cut differently. Caitlin Kalinowski, who had led the company's hardware and robotics division since November 2024, resigned in the days after OpenAI announced it had accepted the Pentagon contract Anthropic had walked away from — citing the company's decision to take the deal as her reason for leaving.
What This Means for AI Tools You Rely On
The immediate business impact of the designation is concentrated in the government contracting space. Defense contractors have six months to transition away from Anthropic products, which in practice means rebuilding applications and workflows built on Claude's API.
But the broader implications extend further than the DoD supply chain.
The case is the first time a major AI company has refused a government demand for unrestricted model access, publicly explained why, and then faced formal government retaliation. Fortune's analysis described it as a turning point: "What just happened with Anthropic and the Pentagon terrifies me" — the concern being that if the government can effectively blacklist an AI company for declining to remove its safety policies, every AI company now faces a choice between ethics and federal contracts.
For businesses that rely on Claude, Claude Code, or any Anthropic API product, the more immediate concern is stability: will the ongoing litigation affect product availability? The short answer, based on current evidence, is no — Anthropic's products remain available to commercial customers and the case is focused on government contracts specifically.
But the case raises a structural question that every company building on AI infrastructure will eventually have to answer: what happens when your AI vendor's policy decisions put them in conflict with government actors? The Anthropic situation is the first live test of that scenario at scale.
What Anthropic's Redlines Actually Represent
It is worth being direct about the substance of what Anthropic refused to enable.
Fully autonomous weapons systems — weapons that select and engage targets without a human making the final decision — are a category that arms control experts, military ethicists, and AI safety researchers have flagged as among the highest-risk applications of advanced AI. The International Committee of the Red Cross has called for legally binding rules restricting them. Multiple United Nations bodies have debated bans.
Mass domestic surveillance — deploying AI to monitor the communications, locations, or behavior of U.S. citizens at scale without individual legal basis — is the kind of capability that civil liberties organizations have warned about for decades.
These are not arbitrary restrictions. They are the two categories where the gap between "what AI can do" and "what we have policy frameworks to safely govern" is widest. Anthropic's position is that it will not fill that gap unilaterally by simply providing the capability and leaving the policy question to someone else.
The Pentagon's position is that an AI vendor cannot put conditions on what lawful government use looks like.
Both positions are coherent. The court will decide which one prevails.
The Bottom Line
Anthropic signed a $200 million Pentagon contract in July 2025. When the government demanded removal of safety restrictions on autonomous weapons and mass surveillance, Anthropic refused. The Trump administration designated them a supply chain risk — an unprecedented move against an American company — and OpenAI moved into the contract hours later.
A federal judge found the designation constituted First Amendment retaliation. An appeals court allowed it to proceed anyway. The case is now in DC court.
What is at stake is not just one contract. It is the question of whether AI companies can maintain meaningful limits on how their technology is used, or whether those limits become negotiating chips the moment a large enough customer demands their removal.
That question does not have a court ruling yet. It will.
Sources:
- Anthropic — "Where things stand with the Department of War" (official statement)
- CNN — Anthropic rejects Pentagon offer: 'We cannot in good conscience accede to their request' (February 26, 2026)
- CNN — Trump orders military contractors to cease business with Anthropic (February 27, 2026)
- AI Thinker Lab — Anthropic Rejects Pentagon AI Contract: Full Timeline (2026)
- Mayer Brown — Pentagon Designates Anthropic a Supply Chain Risk: What Government Contractors Need to Know
- Fortune — OpenAI sweeps in to snag Pentagon contract after Anthropic labeled supply chain risk (February 28, 2026)
- Washington Post — Anthropic sues Pentagon over being labeled a national security risk (March 9, 2026)
- CNBC — Anthropic wins preliminary injunction in DOD fight as judge cites 'First Amendment retaliation' (March 26, 2026)
- CNBC — Anthropic loses appeals court bid to temporarily block Pentagon blacklisting (April 8, 2026)
- CNBC — Anthropic and U.S. government face off in DC court over blacklisting (May 19, 2026)
- Fortune — I've been studying Big Tech for a long time. What just happened with Anthropic and the Pentagon terrifies me (May 16, 2026)
- Fortune — Google and OpenAI employees back Anthropic in its legal fight with the Pentagon (March 10, 2026)
If you're relying on AI tools in your business stack and want to understand the policy and stability risks, book a free call with Sciensify to talk through your options.
