This is not normal.

Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell called an emergency meeting with the CEOs of America's largest banks this week. The agenda: Anthropic's latest AI model, Mythos, and the cybersecurity threat it poses to the entire financial system. Bloomberg first reported the meeting, and Reuters confirmed it Thursday with two sources familiar with the matter.

The meeting, held Tuesday at the Treasury Department in Washington, came while most bank CEOs were already in town for a separate financial conference. Bessent and Powell used the opportunity to deliver a blunt warning: Mythos can identify and exploit vulnerabilities across every major operating system and every major web browser. That is Anthropic's own description of what the model can do.

Think about what just happened here. The two most powerful financial officials in America pulled Wall Street into a room to talk about a product made by an AI startup. Not a foreign adversary. Not a rogue state. A San Francisco company that the Pentagon is simultaneously trying to blacklist.

Anthropic launched Mythos earlier this week as the centerpiece of Project Glasswing, a cybersecurity initiative that gives select companies early access to the model's vulnerability-hunting capabilities. Apple, Google, JPMorgan, and others signed up. Anthropic deliberately stopped short of a broad release, saying the model's ability to find zero-day exploits was too dangerous for open deployment.

That restraint clearly was not enough to calm Washington.

The irony is suffocating. The same government that labeled Anthropic a supply chain risk and kicked it out of the Pentagon is now holding emergency meetings because the company built something too powerful. Anthropic is simultaneously too dangerous to trust with military contracts and so dangerous that the Treasury Secretary needs to warn Wall Street about it. Pick one.

This is the moment the AI safety conversation stopped being theoretical. We are no longer debating hypothetical superintelligence risks in academic papers. The Fed Chair and the Treasury Secretary are telling bank CEOs to shore up their defenses against an AI model that exists right now. The financial system's cybersecurity posture just became an AI problem, and nobody in Washington seems to have a plan beyond calling a meeting.

What happens next matters more than what happened Tuesday. If Mythos can find zero-days across every major platform, it is only a matter of time before a similar model, with fewer guardrails, ends up in the wrong hands. Anthropic chose restraint. The question is whether the next lab will too.