Last week, before most people had heard of Anthropic's Mythos model, Vice President JD Vance and Treasury Secretary Scott Bessent got on a call with some of the most powerful people in technology. On the line: Dario Amodei (Anthropic), Sam Altman (OpenAI), Sundar Pichai (Google), Satya Nadella (Microsoft), Elon Musk (xAI), George Kurtz (CrowdStrike), and Nikesh Arora (Palo Alto Networks).

The topic: what happens when an AI model gets good enough to find vulnerabilities faster than humans can patch them. The answer, apparently, is you call an emergency meeting.

CNBC confirmed that the call focused on two specific concerns: the security posture of large language models and how to respond if AI capabilities begin to scale in ways that give attackers an advantage over defenders. Anthropic's Mythos triggered the call.

Then Came the Bankers

The tech CEO call was just the warm-up. On Tuesday, Bessent and Fed Chair Jerome Powell summoned Wall Street bank CEOs to Treasury headquarters in Washington for a flash meeting. Bloomberg, Reuters, CBS News, Fortune, and CNBC all independently confirmed the meeting. The message to banks: Mythos and models like it represent a new class of cybersecurity threat. Start preparing.

Let that land for a second. The Treasury Secretary and the Fed Chair called in bank CEOs to warn them about a single AI model. Not a state-sponsored cyberattack. Not a market crash. An AI model built by a startup in San Francisco that the Pentagon is currently trying to blacklist.

The Irony Is Deafening

Here is the part nobody is saying out loud. The same government that blacklisted Anthropic from Pentagon contracts just called an emergency meeting because Anthropic's technology is too powerful to ignore. The Pentagon says Anthropic is a supply chain risk. The Treasury Department says Anthropic's model is a national security concern that requires Wall Street's attention. Both cannot be true in the way each side is framing it.

What is actually happening is simpler and scarier: Anthropic built something that exposed how unprepared every institution in America is for AI-powered cybersecurity threats. The government's response was not a plan. It was two separate panic calls in one week. One to the people building these models. One to the people whose systems will be attacked by them.

What Happens Next

Anthropic is only releasing Mythos to select companies through Project Glasswing, giving Apple, Google, Microsoft, and JPMorgan a 90-day window to find and patch vulnerabilities before broader release. That is the most responsible thing any AI company has done. It is also an admission that the model they built can do things that terrify even its creators.

The question is not whether AI-powered cyberattacks are coming. The question is whether two emergency phone calls in a week count as a strategy. Based on what we know so far, the answer is no.

First reported by CNBC. Confirmed by Bloomberg, Reuters, CBS News, and Fortune.