The Court Battle Got Worse for Anthropic

Three Republican-appointed federal appeals court judges heard oral arguments Tuesday — and according to The Hill, Anthropic faced an uphill climb.

Government lawyers argued the Pentagon has broad authority to designate supply-chain risks and cut business relationships without judicial interference. The judges' questions reportedly gave little comfort to Anthropic's legal team.

This follows the appeals court's earlier denial of Anthropic's emergency stay request — which we covered previously. The merits hearing is now on the record, and the panel's skepticism suggests the blacklist survives at least through the appellate process.

Anthopic is fighting to reverse a designation the Pentagon has never used against an American company before in its entire history.

The Pentagon Moved On — Fast

CNN reported the Department of Defense formally announced AI agreements with eight major tech companies: OpenAI, Google, Microsoft, Nvidia, Amazon Web Services, Oracle, SpaceX, and Reflection.

NOT included: Anthropic.

Until recently, Claude was the only AI model operating inside the Pentagon's classified networks. Now Anthropic's competitors have the contracts, the access, and the government money — while Anthropic is stuck in court.

The One Big Beautiful Bill Act allocated a massive sum for Pentagon AI and offensive cyber spending. Tech companies have been lining up for that money. Anthropic is locked out of the room while the table gets set.

The Pentagon said 1.3 million DoD personnel have used its GenAI.mil platform. That's a real user base. Anthropic built it. Someone else will feed it going forward.

What CNN and The Guardian Are Getting Wrong

Left-leaning coverage — CNN and The Guardian especially — keeps framing this as a noble stand against autonomous weapons and mass surveillance.

What's missing from that account: the Pentagon's "lawful operational use" language covers a broad range of legitimate military applications that have nothing to do with killer robots. Logistics. Intelligence analysis. Battlefield communications. Medical triage support.

Anthopic's insistence on embedding specific safety guardrails into DoD contracts — guardrails Anthropic itself defines — means a private company is trying to unilaterally dictate terms of use to the United States military. That's a legitimate policy problem, not just a PR battle.

The Wall Street Journal editorial board noted the national security angle directly: China is not putting ethical guardrails on its military AI. America's adversaries don't pause for corporate safety frameworks. That context is consistently absent from sympathetic coverage of Anthropic.

What Right-Leaning Coverage Gets Wrong

Framing this purely as Anthropic being anti-American is also too simple.

Dario Amodei's company has genuine concerns about autonomous lethal systems operating without human oversight. Those aren't fringe positions — they're shared by serious defense analysts and former military officials across the political spectrum. Pete Hegseth calling it "arrogance and betrayal" is political theater, not strategic analysis.

The real issue is a contract dispute that escalated into a constitutional standoff — and nobody's hands are entirely clean.

Karpathy to Anthropic: The Talent Signal

Almost every outlet buried this detail: Andrej Karpathy joined Anthropic this week.

According to TechCrunch, Karpathy — who co-founded OpenAI, led Tesla's Full Self-Driving program, and is one of the most respected AI researchers alive — announced Tuesday he's joining Anthropic's pre-training team under team lead Nick Joseph.

"I think the next few years at the frontier of LLMs will be especially formative," Karpathy posted on X.

Anthopic confirmed he will build a team focused on using Claude to accelerate pre-training research itself — essentially using AI to make AI smarter, faster.

Pre-training is the most expensive, compute-intensive phase of building frontier models. It's where the core intelligence of a system like Claude gets baked in. Putting Karpathy there is not a symbolic hire. It's a statement about strategy.

Also hired this week: Chris Rohlf, a 20-year cybersecurity veteran who previously worked at Yahoo's "The Paranoids" security team and spent six years at Meta, joining Anthropic's frontier red team to stress-test models against severe threats.

Two serious hires announced the same week the company is hemorrhaging government contracts.

What This Means for Regular People

If you use Claude for anything — work, writing, research — the company behind it is under serious financial pressure. Losing Pentagon contracts is not cosmetic. It's revenue, credibility, and leverage.

If you're a taxpayer, you're funding AI deals with OpenAI, Google, and Microsoft through the Pentagon. Those companies now have access to classified networks and government data that shapes national security decisions. You should know that.

If you think AI safety is just a tech industry marketing term, Karpathy's move to Anthropic is a signal worth paying attention to. The people who understand these systems best are placing their bets. This week, one of the best just bet on Anthropic.

The legal ruling from the appeals court could come any time. When it lands, it determines whether the first-ever blacklisting of an American company by the Pentagon holds — or gets thrown out. Either outcome rewrites the rules.