Ex-security officials urge full appeals court to rethink Anthropic Pentagon ruling
Seventeen former national security officials say a court decision backing the Pentagon's label on Anthropic could chill AI safeguards and federal AI buying.
The short version
- 17 former senior national security officials filed a brief on Oct. 20 asking for a rehearing.
- A D.C. Circuit panel ruled 2-1 against Anthropic's challenge under a 2018 supply chain security law.
- The dispute began when Anthropic objected to mass surveillance and fully autonomous weapons uses of Claude.
- The officials say Hegseth announced the designation before the required review; the formal order came March 3.
- A California federal court ruled for Anthropic on Aug. 27, creating a split picture.
Why it matters
If the ruling stands, the government gets wide room to blacklist a technology supplier after a dispute over how its product may be used. That could change how AI companies negotiate safeguards with federal buyers, and how agencies and contractors plan which models they rely on.
What the former officials are asking
As MeriTalk reported, a group of 17 former senior national security officials has asked the full U.S. Court of Appeals for the D.C. Circuit to take another look at a recent decision. That decision, issued by a three-judge panel, upheld the Pentagon's move to label Anthropic a supply chain risk. The group filed its brief on Oct. 20.
The signers come from both Republican and Democratic administrations and have worked in defense buying, intelligence, cybersecurity and policy. They include Avril Haines, who led national intelligence; Jake Sullivan and Susan Rice, who each served as national security adviser; Wendy Sherman, a onetime deputy secretary of state; and Suzanne Spaulding, who held a senior Homeland Security post.
Their core argument is that the label was driven by politics and that the Pentagon skipped steps the law requires. They also warn the ruling could become a template for punishing companies that fall out of favor.
How the fight started
The conflict began when the Pentagon wanted to use Anthropic's Claude model for any lawful purpose. Anthropic objected to two possible uses: mass surveillance and fully autonomous weapons, meaning weapons that act without a person making the final call.
Talks broke down. Defense Secretary Pete Hegseth then directed the department to designate Anthropic a supply chain risk. Anthropic sued, saying the move was unlawful retaliation. The Trump administration defended it as a national security step.
The panel sided with the government by a 2-1 vote. It rejected Anthropic's challenge under the Federal Acquisition Supply Chain Security Act of 2018, a law that lets the government deal with supply chain risks in federal purchasing. The judges found the Defense Department had solid grounds to conclude that using Claude, whether by the department or its contractors, posed a national security risk.
The case that process was backwards
The former officials say a real risk finding follows a sequence. An agency names a specific threat, weighs the possible consequences, considers alternatives, and writes down its reasoning, all before acting. In their account, the Pentagon did none of that in time.
They point to the calendar. Hegseth announced the designation publicly, and the formal determination came on March 3, four days later. The brief argues that reasons were supplied afterward, and says that conclusion first and justification later is the hallmark of pretext, according to the officials. Pretext means a stated reason that hides the real one.
The filing also argues the government violated Anthropic's First Amendment free-speech rights. It cites Hegseth's criticism of Anthropic's rhetoric, which he called sanctimonious and virtue-signaling, as evidence that the company's public views on AI safety shaped the decision.
Why they say the stakes go beyond one company
The brief makes two broader warnings. The first is about access. As the U.S. competes with China on technology, the officials say shutting a leading AI developer out of federal buying could cost the government useful capabilities. They note that the Pentagon has continued to show interest in Claude Mythos, an Anthropic model with advanced cybersecurity abilities, as an example of what exclusion could cost.
The second is about incentives. If any safeguard the Pentagon dislikes can trigger a risk label, the officials say, companies cannot credibly promise to build products safely. That could change how other AI developers bargain with federal agencies over how their tools are used.
The newsletter's earlier framing holds: this is an argument from named critics, not a settled outcome. The full court has not yet said whether it will rehear the case.
A split in the courts
The brief also leans on a separate case. On Aug. 27, a federal district court in Northern California gave Anthropic a summary judgment, a ruling made without a full trial, on its claims of First Amendment, due process and Administrative Procedure Act violations. The Administrative Procedure Act sets the rules agencies must follow when making decisions.
So one court has sided with Anthropic on the designation, while the appeals panel in Washington has backed the Pentagon under a different law. The former officials are asking the full D.C. Circuit to reconsider the panel's ruling.
What to watch
- Will the full D.C. Circuit agree to rehear the case?
- How will the D.C. Circuit panel ruling and the California district court ruling be reconciled?
- Will other AI developers change how they negotiate usage limits with federal agencies?
Try it
If you work in government contracting or advise public-sector clients, ask your compliance team whether any current project names Anthropic models, and note what switching would involve.
Sources and links
- ReportingMeriTalk: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by MeriTalk · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
