The takeaway
The decision pits two federal courts against each other and pushes the fight over AI model restrictions toward the Supreme Court, with direct consequences for how frontier labs control deployment.
Why it matters for builders
Frontier AI labs are learning that safety guardrails cut both ways: the same restrictions that build trust can be read as a national-security liability. Builders shipping models to government or defense customers should expect legal scrutiny over model behavior, not just benchmark performance. Watch this case for the precedent it sets on whether a lab can refuse a customer on the basis of its own safety policy.
Court Upholds Pentagon's Blacklisting of Anthropic Over Claude
A US appeals court has sided with the Pentagon in its months-long fight with Anthropic, ruling that the government can blacklist the AI company for refusing to enable certain Claude features for military use. The 2-1 decision from the US Court of Appeals for the District of Columbia Circuit marks a major escalation in the battle over how "safe" AI models should behave in national-security contexts.
What happened
In a ruling issued September 25, a three-judge panel denied Anthropic's petitions for review, clearing the way for the Department of Defense to keep barring federal agencies from using Anthropic's products and banning defense contractors from doing business with the company. The dispute dates back to March, when the Trump administration ordered the blacklisting after Anthropic refused to drop restrictions on Claude's use for lethal autonomous warfare and mass surveillance.
The judges framed the case as a collision between two "deeply sobering" risks. On one side, "overly constrained AI models" could shut down unexpectedly and cause military operations to fail. On the other, "unconstrained AI models" might hallucinate "inappropriate targets for lethal military force." The court concluded that Defense Secretary Pete Hegseth "must determine how best to balance the competing risks" and did not overstep his authority under the Supply Chain Security Act.
The two judges who ruled against Anthropic, Gregory Katsas and Neomi Rao, were both appointed by Trump. Anthropic said it "respectfully disagrees" with the decision and is "considering all options, including further review," potentially through an en banc appeal or the Supreme Court.
Why it matters
The ruling directly conflicts with a separate decision from a California federal court, where Judge Rita Lin found the blacklisting violated the First Amendment and that Anthropic does not meet the statutory definition of a "supply chain risk." That split makes a Supreme Court showdown increasingly likely.
For AI builders, the case crystallizes a question with no clean answer: how much behavioral restriction should model providers be allowed to encode into their products, and who gets to decide? Anthropic's safety guardrails, which it markets as a feature, are here being treated by the government as a defect that threatens national security. Whichever way the courts ultimately land, the outcome will shape how much control frontier labs retain over how their models are deployed, and by whom.
The Automation Brief
Read 5 AI stories instead of 50.
The essential moves in AI agents, models, automation and infrastructure — filtered for builders and operators, with the part that actually matters.
No noise. Unsubscribe anytime.
Editorial notes
Stefan Trbojevic
n8n Lab Editorial
26 September 2026
26 September 2026
Sources
AI disclosure: AI assisted with research and drafting. Factual claims are reviewed by an editor.
