Appeals court backs Pentagon blacklist

A federal appeals court ruled the Trump administration can blacklist Anthropic for refusing to remove safety restrictions from Claude, even without proving the company acted maliciously.

The core tension

The court acknowledged the dilemma: overly restricted AI could fail military operations, but unrestricted AI could target wrong objectives for lethal force. The Pentagon must balance these competing risks.

“the US raises the deeply sobering prospect of overly constrained AI models shutting down unexpectedly and thus causing important military operations to fail”

Two judges, Trump connection

The two judges who ruled against Anthropic were both appointed by Trump. Judge Gregory Katsas was deputy counsel to the president, and Judge Neomi Rao worked in Trump's Office of Management and Budget.

Conflicting court decisions

A California federal judge ruled last month that the blacklist was illegal because Anthropic doesn't meet the definition of a supply chain risk. Today's DC Circuit decision disagreed, citing different legal authority.

Different laws, different outcomes

The California court reviewed the blacklist under a law requiring proof of malicious intent. The DC Circuit reviewed it under a broader law that doesn't require bad motive, giving the Pentagon more authority.

Bad motive not required

The appeals court said the broader law covers any person who denies or manipulates technology function, not just adversaries acting with malicious intent. Anthropic's refusal to enable features qualifies as denial.

“the Secretary's concern about Anthropic disabling Claude from performing lawful actions requested by the Department qualifies as a supply chain risk”

Pentagon's specific concern

The Defense Department worried that Anthropic might prevent Claude from performing national security functions the military deemed necessary, including tasks for an overseas military operation.

One judge dissented

Judge Karen Henderson, a George H.W. Bush appointee, disagreed. She said the law targets hostile actors infiltrating supply chains, not contractors honestly enforcing their own restrictions on product use.

“a contractor's honest and upfront enforcement of restrictions on a covered article's use disfavored by the government”

Anthropic's restrictions

Anthropic encodes limits into Claude to prevent use in lethal autonomous warfare and mass surveillance. The company says the Trump administration retaliated after it refused to drop these ethical restrictions.

California judge's view

The lower court judge ruled the blacklist violated the First Amendment, writing that invoking national security cannot be a blank check to punish government critics.

“The empty invocation of national security is not a blank check to punish and retaliate against government critics”

What happens next

Anthropic says it disagrees with the ruling and is considering further appeals, either to the full DC Circuit or the Supreme Court. Commerce Secretary Howard Lutnick recently claimed the two sides have patched things up.