Emergency Meeting on Mythos Risk

On Tuesday in Washington, Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell reportedly pulled the chief executives of major Wall Street banks into a closed-door meeting about one artificial intelligence model: Anthropic’s Mythos. Reuters and Bloomberg both reported that the warning centered on cyber risks, not market prices or bank balance sheets. (finance.yahoo.com) (bloomberg.com) Anthropic unveiled Claude Mythos Preview on April 7, 2026, and called it its “most capable frontier model to date.” In its own system card, the company said the jump showed up most sharply in computer security work. (anthropic.com) Computer security work here means finding holes in software before criminals do, like spotting an unlocked side door in a skyscraper before anyone tests the handle. Anthropic said Mythos was unusually strong at finding and exploiting “zero-day” vulnerabilities, which are software flaws that defenders have not patched yet because they do not know they exist. (anthropic.com) Anthropic did not throw the model open to the public. The company said it would not make Mythos Preview generally available and instead created Project Glasswing, a limited program to use the model for defensive security work. (anthropic.com 1) (anthropic.com 2) That detail explains why bank regulators got involved so fast. If a model can help a defender scan code faster, the same capability can also help an attacker map weak points across payment systems, trading infrastructure, cloud services, and vendor software that large banks rely on every day. (anthropic.com) (bloomberg.com) Banks already treat cyberattacks as an operational risk, which is the plain-English term for a failure in the machinery that keeps money moving. The Federal Reserve and the Treasury Department stepping in together suggests officials are treating advanced artificial intelligence less like a normal software launch and more like a possible stress event for core financial plumbing. (bloomberg.com) (finance.yahoo.com) The reported meeting was not about a hack that had already happened. Reuters said the purpose was to make sure banks understood the future risks raised by Mythos and similar models and were taking precautions before those capabilities spread further. (finance.yahoo.com) Anthropic says more than 99 percent of the vulnerabilities it found during testing have not yet been patched, which is why it withheld technical details. That is the kind of sentence that gets attention in a bank boardroom, because one unpatched flaw inside a widely used vendor can ripple through dozens of institutions at once. (anthropic.com) Project Glasswing started with a small set of launch partners and Anthropic said it also extended access to more than 40 organizations that build or maintain critical software infrastructure. Reuters reported that access to Mythos would stay limited, which means the immediate fight is to harden systems before tools with similar abilities become easier to obtain. (anthropic.com) (finance.yahoo.com) The shift here is simple and new: top financial regulators are now reacting to an artificial intelligence model the way they would react to a dangerous weakness in the pipes under the street. When the Treasury secretary, the central bank chair, and the biggest bank chiefs all clear their calendars for the same cyber call, the model is no longer just a tech story. (bloomberg.com)