US Warns Banks: AI Poses Cyber Threat

U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell met this week with chief executives from major U.S. banks to discuss cybersecurity risks tied to a newly launched artificial intelligence model from Anthropic, according to officials familiar with the meeting.

 

The session took place on Tuesday at the Treasury Department in Washington, D.C., and was framed as a warning to financial institutions about potential vulnerabilities and the need to ensure defensive measures are in place. Officials described the engagement as focused on protecting critical financial infrastructure as AI capabilities advance quickly.

 

The discussion centered on Anthropic’s new model, Mythos, which the company launched earlier this week. Anthropic limited the model’s broad release after raising concerns about previously unknown cybersecurity vulnerabilities, the company said. Anthropic also stated that Mythos is capable of identifying and exploiting weaknesses across all major operating systems and web browsers.

 

Access to Mythos is currently restricted to approximately 40 technology companies, including Microsoft and Google, according to the company. The restricted rollout was cited as part of the company’s approach to managing the risks it identified around the model’s potential cybersecurity impact.

 

Bank leaders attending the meeting included the CEOs of Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs. JPMorgan CEO Jamie Dimon was reportedly unable to attend.

 

is a direct, high-level outreach by U.S. financial authorities to the largest banks about a specific AI-related cybersecurity concern linked to a named model and developer. What it means is that regulators are treating fast-moving AI capabilities as a potential risk factor for the resilience of the financial system’s digital backbone, particularly where vulnerabilities could affect widely used operating systems and browsers.

 

For global markets, the meeting highlights how AI deployment decisions by a single technology provider can become a topic for financial supervisors when the potential exposure touches core banking operations. For global politics and cross-border regulation, it underscores that U.S. authorities are engaging directly with systemically important institutions on technology risks that can propagate internationally through interconnected networks and shared software dependencies.

 

Key unknowns remain, including the specific nature of the vulnerabilities Anthropic referenced and how broadly similar capabilities may exist in other advanced AI systems. Officials did not publicly detail any required actions beyond ensuring defensive measures are in place, and the company’s restrictions on access indicate ongoing risk management around the model’s release.