US banking regulators have temporarily paused some cybersecurity examinations for the nation’s largest banks, a move that gives lenders breathing room to address a vast number of software flaws exposed by Anthropic’s new AI model, Mythos. The decision by the Federal Reserve and the Office of the Comptroller of the Currency (OCC) shows how quickly advanced AI has shifted the balance of power in the cybersecurity arms race.
"It’s serious work. We have, I think, hundreds of people doing it full time now," JPMorgan Chase & Co. CEO Jamie Dimon said at an Anthropic event in May, confirming the scale of the banking industry's response.
The regulatory pause comes just weeks after Anthropic’s April rollout of Mythos, a frontier AI model that demonstrated an unprecedented ability to identify zero-day exploits across major operating systems and browsers. In a closed-door briefing in Washington, Treasury Secretary Scott Bessent and then-Fed Chair Jerome Powell warned CEOs from JPMorgan, Goldman Sachs, Morgan Stanley, and other systemic lenders about the potential for the technology to be weaponized. Under an initiative called Project Glasswing, a handful of firms were granted early access to Mythos to test their own defenses, uncovering thousands of vulnerabilities.
For Wall Street and its overseers, the episode has become a race against time. Anthropic CEO Dario Amodei has publicly warned of a six- to 12-month window for financial institutions to patch the tens of thousands of flaws identified by Mythos before rival AI models with similar capabilities are developed, potentially by state actors in China. Fed Vice Chair for Supervision Michelle Bowman said that while exams are paused, regulators will "continue to focus on critical developments and communicating these risks to supervised institutions." The OCC is reportedly conducting its own trials with the model to better understand its capabilities before resuming examinations.
This article is for informational purposes only and does not constitute investment advice.