admin
The D.C. Circuit just handed the Pentagon a clear win in the fight over who controls military AI. The appeals court upheld the Department of War’s decision to bar Anthropic’s Claude from defense systems under the Federal Acquisition Supply Chain Security Act. This is more than a legal squabble — it reshapes how Washington will buy and trust artificial intelligence tools for national security.
D.C. Circuit backs Pentagon exclusion of Claude
The three-judge panel ruled 2–1 that the Department of War acted lawfully when it excluded Claude from Department systems and from use by contractors under 41 U.S.C. § 4713. The majority said the Pentagon had “ample support” to view Anthropic’s contractual limits as a supply‑chain risk. The court rejected Anthropic’s challenges on procedure and free-speech grounds. Judge Karen LeCraft Henderson dissented, saying the majority stretched the law beyond what Congress meant. In plain terms: the Pentagon warned it could not rely on software that might refuse a lawful military task, and the court sided with the Pentagon.
Why this matters for defense procurement and AI companies
This decision tells defense contractors and AI firms that ethics clauses and product guardrails are not cost-free when it comes to national-security work. If a company decides it will block certain uses — even for reasons that sound noble — the Pentagon can treat that choice as an operational risk and keep the product out of its supply chain. At the same time, a separate California judge reached a different result under another law, so the legal picture is not tidy. But for now, the D.C. Circuit gives the Department broad authority in procurement decisions involving AI.
Anthropic’s ethics vs. the military’s needs
Anthropic froze two big no-nos: no mass domestic surveillance and no fully autonomous lethal weapons. Admirable on its face. Problem is, wartime and battlefield needs do not always line up with tech-company moral pledges. Secretary of War Pete Hegseth pushed for contractors to accept “all lawful uses” so commanders won’t be surprised when the chips are down. Anthropic can keep its conscience. The Pentagon can keep a supply chain it trusts. If you want both, good luck threading that needle.
What’s next: appeals, Congress, and real-world fallout
Expect more litigation. Anthropic says it will consider further review, and the split between the D.C. Circuit and the California district court invites more fights — maybe en banc rehearing or even a Supreme Court bid. Congress could also weigh in to clarify procurement law for AI. For now the ruling is a blunt reminder: national security law will beat corporate virtue signaling when the two collide. If you run an AI shop and hope to sell to the Pentagon, plan for a world where safety PR lines might cost you government work — and where the government will not be mocked for insisting on tools that behave when lives are on the line.
