In the three weeks since the Pentagon designated Anthropic a "supply-chain risk" — a sanction normally reserved for foreign adversaries like Huawei — something extraordinary has happened in Washington's courtrooms. Microsoft has filed a legal brief warning of severe economic damage to the American AI industry. Former CIA Director Michael Hayden and a coalition of retired generals have accused Defense Secretary Pete Hegseth of "retribution" against a private company. Thirty-seven engineers from OpenAI and Google DeepMind have raised their names to call the government's actions an "improper and arbitrary use of power." And in a left-right fusion that would have seemed unthinkable a year ago, the Electronic Frontier Foundation and the libertarian Cato Institute have jointly invoked the First Amendment. Meanwhile, Anthropic itself is playing offense: it launched the Anthropic Institute on March 11, consolidating its safety research under co-founder Jack Clark and opening a Washington, D.C. policy office. This is no longer one company's legal fight. It has become a defining test of who gets to set the rules for artificial intelligence in America.
How the Conflict Began
The dispute traces back to late February, when negotiations between Anthropic and the Department of War broke down. Anthropic CEO Dario Amodei had drawn two red lines in the company's government contract: Claude would not be used for mass domestic surveillance of Americans, and it would not be embedded in fully autonomous lethal weapons systems. The Pentagon's position, according to court filings, was that it should be able to use AI for any "lawful" purpose and that a private contractor had no standing to impose ethical constraints on government operations.
Amodei published a statement on February 26 saying he "cannot in good conscience accede to the Pentagon's request" for unrestricted access. Within days, the DOW used the Defense Federal Acquisition Regulation Supplement to designate Anthropic a supply-chain risk — a designation that prevents the company from receiving government contracts, bars military prime contractors from using Anthropic's technology, and effectively cuts off a company that had been embedded at national nuclear laboratories and throughout classified military networks.
TTN covered the immediate fallout in depth: Anthropic filed dual federal lawsuits on March 9, seeking a temporary restraining order and challenging the constitutional validity of the designation. The same day the lawsuits were filed, the DOD announced a new AI contract with OpenAI — a timing that many in the industry read as deliberate signaling.
Rivals to the Rescue: OpenAI and Google DeepMind Workers
The most striking early development was the speed with which engineers at rival companies came to Anthropic's defense — not as company representatives, but as individuals signing in their personal capacity. Within hours of the lawsuits being filed, more than 37 researchers and engineers from OpenAI and Google DeepMind filed an amicus brief with the court, according to Wired.
The brief's signatories include Jeff Dean, Google DeepMind's chief scientist and one of the most decorated computer scientists in Silicon Valley, along with OpenAI researchers Gabriel Wu and Pamela Mishkin, and Google DeepMind researchers Zhengdong Wang and Noah Siegel. For employees at competing AI companies to publicly side with a rival in a lawsuit against their own government's military was, by any measure, an unusual act of solidarity.
Their argument was not sentimental. "The government's designation of Anthropic as a supply chain risk was an improper and arbitrary use of power that has serious ramifications for our industry," the brief states, as reported by TechCrunch. The engineers argued that the Pentagon had a straightforward option available to it if it was unhappy with Anthropic's terms: cancel the contract. Using a national security designation against a domestic AI company that refused to weaken its safety guardrails, they wrote, "chills open deliberation in our field about the risks and benefits of today's AI systems."
The brief also made a structural argument that carries weight beyond this particular case. Without public law explicitly governing how AI can be deployed by the military, it argues, the contractual and technical restrictions that developers impose on their own systems are the only meaningful safeguard against catastrophic misuse. Strip those away and there is nothing left.
Microsoft and the Economic Stakes
The corporate filings that followed amplified the legal argument into an economic one. Microsoft filed a brief stating that the supply-chain risk designation "may bring severe economic effects that are not in the public interest," according to Euronews reporting on the filings. Microsoft's language was carefully calibrated: the company stopped short of accusing the government of acting in bad faith but made clear that the precedent being set here — using a national security designation to penalize a domestic company for maintaining ethical policies — threatened the entire commercial AI sector.
The argument resonates because of scale. Anthropic had been the only commercial AI company approved for classified military networks. Its Claude model was operational at national nuclear laboratories and across DOW intelligence infrastructure. The sudden reversal, and the designation of that same company as a supply-chain risk within a matter of weeks, is the kind of regulatory whiplash that makes corporate counsel at every AI company in America nervous about the durability of their government relationships.
OpenAI CEO Sam Altman, whose company is now holding the Pentagon contract that Anthropic lost, posted on social media that enforcing the supply-chain risk designation against Anthropic "would be very bad for our industry and our country." The awkwardness of that position — benefiting financially from a situation you've publicly criticized — has not been lost on observers, but Altman's willingness to say it aloud is itself notable.
The Military and Intelligence Establishment Speaks
Perhaps the most surprising voices in the coalition came not from Silicon Valley but from the national security community itself. A filing supported by former CIA Director Michael Hayden, along with a group of retired military officials, accused Defense Secretary Hegseth's conduct of amounting to "retribution against a private company that has displeased the leadership," according to Euronews. The ex-officials characterized the DOW's actions as a misuse of government authority that "threatens the rule-of-law principles that have long strengthened our military."
The retired generals' brief also raised an operational concern that cuts to the heart of the crisis: Anthropic's AI was not peripheral to military operations — it was embedded in them. The sudden "uncertainty" of targeting technology widely integrated into military platforms could, they warned, disrupt planning and put soldiers at risk during ongoing operations. The filing explicitly referenced the war in Iran, where AI-assisted targeting has been in active use. In other words: the Pentagon's decision to punish Anthropic may have undermined operational capabilities that the Pentagon itself had come to rely upon.
The Constitutional Argument: EFF and Cato United
One of the more philosophically significant filings came from a left-right coalition that included the Electronic Frontier Foundation and the Cato Institute — two organizations that agree on almost nothing except the importance of limiting government overreach into private expression and commerce. Their joint brief framed the government's actions as a violation of the First Amendment, arguing that forcing an AI company to remove the constraints it places on the outputs of its own systems is tantamount to compelling speech.
"It is not hard to imagine a world in which the government effectively controls what all Americans do and say," the brief states, if federal agencies can dictate how private AI companies configure their models. "The government's actions threaten the vitality and independence of our democracy." The First Amendment framing, if it gains traction in the DC Circuit, could have sweeping implications — not just for AI but for every technology company that makes editorial or algorithmic choices about what its products will and will not do.
Anthropic Launches the Anthropic Institute
While the legal battle escalates, Anthropic has moved to institutionalize its position. On March 11, the company announced the formation of the Anthropic Institute, a new internal unit consolidating three existing teams: the Frontier Red Team, which studies AI-related cybersecurity risks; the Societal Impacts team, which evaluates how workers use Claude autonomously; and the Economic Research unit, which tracks what business activities are being automated with AI. The Institute will be led by Jack Clark, Anthropic's co-founder, who has been appointed the company's head of public benefit.
The Institute's research roster is notably high-caliber. Matt Botvinick, formerly a senior director of research at Google DeepMind, has joined to lead cognitive science research. Zoë Hitzig, previously at OpenAI, has moved over to connect economics work to model training. Anton Korinek, an economics professor specializing in AI's impact on labor and economic activity, will lead a project studying how AI could reshape the structure of economic activity across sectors.
Simultaneously, Anthropic announced it will open a Washington, D.C. office for its Public Policy team this spring. The timing is pointed: the company that was just designated a national security threat by the Pentagon is now planting a flag in the capital, signaling that it intends to be a permanent, active participant in shaping the regulatory environment it operates in — not a passive subject of it.
Why This Moment Matters Beyond One Lawsuit
It would be easy to read the Anthropic case as a dispute between one company and one administration. But the breadth and composition of the coalition now rallying to Anthropic's defense suggests something structurally deeper is at stake.
The core question being litigated — whether the federal government can use national security law to punish a domestic AI company for maintaining ethical constraints on its own products — has no settled precedent. If the courts rule against Anthropic, the practical consequence is that every AI company in America faces a binary choice when dealing with federal agencies: remove your safety policies or accept designation as a national security threat. That would restructure the entire relationship between the AI industry and the government, giving federal procurement power an effective veto over the ethical commitments of private AI developers.
The fact that this concern has united Microsoft and the EFF, former CIA directors and libertarian think tanks, and engineers at rival companies who are simultaneously benefiting from Anthropic's misfortune — that kind of coalition does not form around narrow corporate interests. It forms when the underlying principle touches something that everyone in an industry depends on: the ability to say what their own products will and will not do.
TTN has tracked the financial dimensions of this conflict closely — court filings reveal Anthropic spent over $10 billion while earning just $5 billion in lifetime revenue, making the stakes of losing government business existential. But the coalition filing briefs in the DC Circuit is not primarily arguing about Anthropic's balance sheet. They are arguing about what kind of AI industry America wants to have — and who gets to decide.
This is part of TTN's ongoing coverage of the Anthropic–Pentagon conflict. See also: Anthropic Files Dual Federal Lawsuits to Block Pentagon's Unprecedented Supply Chain Blacklist.
