Anthropic Showed Trump Their Cards Before He Flipped the Table

Imagine spending weeks preparing a presentation for a client, nailing the pitch, getting positive feedback—then watching them announce on Twitter they’re done with you. That’s essentially what happened to Anthropic with the Trump administration and their AI model Mythos.

According to Anthropic co-founder Jack Clark, speaking at the Semafor World Economy summit, the company briefed the Trump administration on Mythos before the former president declared the relationship over. The timing here is fascinating in the worst possible way. You prep the briefing, you deliver it, and then—surprise—the partnership is supposedly dead.

What We Actually Know

The facts are sparse but telling. Anthropic gave the Trump administration a briefing on Mythos, which the company describes as having powerful capabilities. The exact date of this briefing remains unknown, but it happened before Trump’s public declaration ending the relationship.

Here’s where it gets interesting: a new court filing reveals the Pentagon told Anthropic the two sides were “nearly aligned” just one week after Trump announced the relationship was kaput. So which is it? Are they done, or are they almost aligned? This is the kind of mixed messaging that makes toolkit reviewers like me want to pull our hair out.

The Toolkit Reviewer’s Take

From a practical standpoint, this situation highlights something I see constantly in the AI tools space: the gap between technical capability and organizational reality. Anthropic can build whatever powerful model they want—and by all accounts, Mythos has serious capabilities—but none of that matters if the partnership dynamics are this messy.

When I review AI toolkits for agntbox.com, I’m always looking at whether a tool actually works in real-world conditions, not just in demos. This Anthropic-Trump situation is the enterprise version of that same problem. You can have the best briefing in the world, but if the decision-maker has already mentally checked out, you’re presenting to a wall.

What This Means for AI Partnerships

The fact that Clark is still explaining why Anthropic remains engaged tells you everything. Companies don’t usually need to justify why they’re continuing conversations with potential partners unless those conversations have become publicly awkward. This is damage control dressed up as transparency.

For those of us who evaluate AI tools professionally, this raises questions about how these major AI companies handle government relationships. Are they getting clear commitments before investing resources in briefings? Do they have backup plans when political winds shift? Based on this situation, it seems like even the big players are figuring this out as they go.

The Bigger Picture

What bothers me most about this story is the opacity. We don’t know what Mythos actually does beyond “powerful capabilities”—a description so vague it could apply to anything from a chatbot to a weapons system. We don’t know the briefing date. We don’t know what “nearly aligned” means in Pentagon-speak. We’re left connecting dots with half the picture missing.

This is exactly the kind of situation that makes AI governance so difficult. How are we supposed to have informed discussions about AI deployment in government when the basic facts of who briefed whom and when remain unclear?

Clark’s willingness to discuss this publicly at least shows Anthropic isn’t trying to pretend it didn’t happen. That’s something. But transparency about a confusing situation is still just transparent confusion.

For now, Anthropic appears to be in a strange limbo—briefing an administration that may or may not want to work with them, getting encouraging signals from the Pentagon, and trying to explain it all to a skeptical public. As someone who tests tools for a living, I can tell you this much: uncertainty is the enemy of adoption. And right now, there’s uncertainty everywhere you look.

🕒 Published: April 15, 2026