The Government Loves AI It Doesn't Trust

The contradiction arrived quietly, as the best ones do.

A writer at TechCrunch has reported that Trump officials may be encouraging banks to test Anthropic's Mythos model — this, despite the Department of Defense having recently declared Anthropic a supply-chain risk. Two arms of the same government. Two completely opposite postures toward the same company. Both, apparently, official.

This is not a bug. This is the operating system.

The Right Hand, the Left Hand

The thing about watching the federal government engage with AI is that you start to notice it doesn't have a position — it has positions, plural, held simultaneously by different offices with different agendas and no apparent obligation to reconcile them. The DoD flags Anthropic as a liability. Somewhere else in the building, officials are nudging banks to open accounts with the same technology. Nobody seems to have cc'd each other on the memo.

What TechCrunch has surfaced here isn't really a scoop about Anthropic. It's a scoop about institutional incoherence. The fact that this contradiction exists — and exists openly enough that it can be reported — tells you something important: the government hasn't decided what AI is yet. Is it infrastructure? Is it a foreign-policy-adjacent risk? Is it a productivity tool for financial services? The answer, right now, is yes to all of them, depending on which floor you're on.

Anthropics's position in this is almost beside the point. The company is just the object the confusion is being projected onto.

What the Cycle Looks Like From Here

I've watched enough technology go through the Washington gauntlet to recognize the shape of this moment. First comes the enthusiasm — someone high up gets a demo and wants it everywhere. Then comes the caution — a different someone reads a briefing and wants it nowhere. Then, eventually, comes the framework, usually years late and written by people who understand neither the technology nor the problem it was supposed to solve.

We are somewhere in the middle stretch right now. The enthusiasm and the caution are happening at the same time, in the same administration, about the same product. That's not unprecedented. But it is clarifying.

What it clarifies is this: the government's relationship with AI isn't being managed. It's being improvised. Different agencies are running different plays with no shared playbook, and the result is that a company can be simultaneously encouraged and flagged — a supply-chain risk worth piloting in your bank.

For Anthropic, there's probably a version of this that works out fine. Governments have a long history of using things they're officially nervous about. But for anyone trying to read Washington's position on AI as a signal — investors, foreign governments, competing labs — the signal right now is noise. Loud, contradictory, bipartisan noise.

The most honest thing you can say about where the U.S. government stands on AI is that it doesn't know where it stands, and it's doing two things at once to prove it.