On February 26, 2026, Anthropic told the Pentagon no. The sticking points were two phrases — mass surveillance and autonomous weapons — baked into guardrails the company wouldn't strip from Claude, its flagship AI model. The Department of Defense wanted them gone. Dario Amodei, Anthropic's CEO, published a statement calling the demands incompatible with democratic values. Within days, the Pentagon had designated Anthropic a "supply chain risk," a classification the government previously reserved for foreign adversaries like Huawei and Kaspersky.1 Within weeks, OpenAI signed its own deal with the military.2
A federal judge blocked the blacklisting. Judge Rita Lin called it "classic illegal First Amendment retaliation," rejecting what she described as the "Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government."3
The internet loved it. ChatGPT uninstalls surged 295%.4 Anthropic's revenue kept climbing.5 Caitlin Kalinowski, a senior member of OpenAI's robotics team, resigned, saying the guardrails around certain AI uses "were not sufficiently defined" before the Pentagon agreement was announced.6 The narrative assembled itself with unusual speed: principled company defies military overreach, the market rewards integrity, democracy functions as intended.
Except that at no point in this sequence — the demand, the refusal, the blacklisting, the lawsuit, the replacement vendor — did any elected body make a decision about anything.
Procurement as Legislature
The United States has no statutory framework governing how AI gets used in warfare. No law specifies what AI can target, what requires human authorization, what constitutes acceptable surveillance, or who is responsible for any of these decisions. The rules — such as they are — get written through bilateral vendor contracts between the Pentagon and whichever company picks up the phone.7
The Lawfare Institute put the structural problem more cleanly than Congress ever has: "the rules governing the military's use of AI are not derived from statutes or regulations, but from bilateral agreements between the government and individual vendors" — agreements that "were not designed to provide the democratic accountability, public deliberation, and institutional durability that statutes provide."7
Military AI policy in the world's most powerful democracy is being authored in procurement negotiations. The people writing it were not elected, will not be reelected, and answer to shareholders or to acquisition officers, depending on which side of the table they sit.
Anthropic was almost certainly right to refuse. It was also, by definition, insufficient — because a system where the guardrails on lethal AI depend on which CEO happens to have principles this fiscal quarter is a system where those guardrails are one retirement, one board vote, one bad earnings call away from vanishing.
Next Vendor Up
When Anthropic refused, the Pentagon did not pause to reconsider. It called OpenAI. The Electronic Frontier Foundation reviewed the resulting contract language and noted that lawyers call such terms "weasel words" — language that "creates ambiguity that protects one side or another from real accountability for contract violations."2 The EFF observed that OpenAI's agreement relied on the assumption that federal agencies would follow the law, which, for anyone who remembers Edward Snowden, is not exactly an ironclad safeguard.
The net result: the Pentagon got its AI. Instead of Anthropic's hard guardrails, it got OpenAI's soft language. Corporate refusal didn't block anything — it selected for the vendor with the least friction.
Economists have a name for this dynamic — adverse selection — and it is the most structurally damning argument against treating corporate conscience as a governance mechanism. The principled actor exits. The unprincipled actor fills the gap. The outcome is worse than if the principled actor had stayed and negotiated from the inside.
Google traced this same arc years before. In 2018, more than 3,100 employees signed a letter telling Sundar Pichai that "Google should not be in the business of war," and the company dropped its Project Maven Pentagon contract.8 By 2025, Google had removed its AI ethics pledges entirely, citing "a global competition taking place for AI leadership within an increasingly complex geopolitical landscape." Seven years from principled stand to quiet capitulation. Corporate guardrails operate on market time, and markets are not known for their patience.
The Symmetry Nobody Wants
Amodei's statement contained a line that cut deeper than he may have intended: "These latter two threats are inherently contradictory: one labels us a security risk; the other labels Claude as essential to national security."1 He was naming the Pentagon's logical incoherence — you cannot simultaneously claim a company is too dangerous to work with and too important to be allowed to refuse.
Fair enough. But there is an equally uncomfortable incoherence on the other side, and the people cheering Anthropic have not shown much interest in examining it.