Treasury and Fed warn big banks about cyber risks from Claude Mythos

In short: US Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell held a last minute meeting with top Wall Street bank CEOs to warn about cyber risks tied to Anthropic’s AI model, Claude Mythos.

What happened

The meeting took place on April 10, 2026, and was first reported by Bloomberg. It included leaders from large banks such as Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs. These are often called “systemically important” banks, meaning problems at one of them can ripple out and affect the wider financial system.

Officials urged the banks to strengthen their cyber defenses because of concerns about Claude Mythos. According to reports, the model can identify and exploit software weaknesses across major operating systems and web browsers. A software weakness, also called a vulnerability, is like an unlocked side door in a building that most people do not know is there.

One worry discussed was that Mythos may be better than most humans at finding these hidden doors and figuring out how to use them. Anthropic has said the model has uncovered thousands of previously unknown vulnerabilities, including one that was more than 25 years old. The fear is that an attacker could use an AI system like this to automate complex cyberattacks at large scale, including efforts to break encryption (the digital equivalent of a locked safe) or disrupt critical services like banking.

Why it matters

Banks hold money, personal data, and payment systems that millions of people rely on every day. If AI tools make it easier to find and exploit weak spots in software, the risk of large hacks and service outages could increase, even for well defended institutions.

Source: NYTimes