The US Treasury secretary, Scott Bessent, has reportedly met with major American bank leaders this week as officials assessed potential cyber threats that Anthropic’s latest artificial intelligence system poses.

Summary

• Scott Bessent convened major U.S. bank CEOs to assess cybersecurity risks linked to Anthropic’s Claude Mythos AI model following a code leak.
• The model reportedly uncovered thousands of long-standing software vulnerabilities, raising concerns over misuse by hackers and threats to financial stability.
• Anthropic’s revenue surpassed $30 billion annualized, driven by enterprise demand, major compute deals with Google and Broadcom, and the growth of its Claude Code platform.

According to reports, Treasury Secretary Scott Bessent brought together senior executives at the department’s Washington headquarters, with Jerome Powell also said to be present. The meeting followed the unveiling of Anthropic’s Claude Mythos model, which the company has described as posing “unprecedented” cybersecurity risks.

Concerns surrounding the model intensified after its code was leaked earlier this month. In a subsequent blog post, Anthropic said advanced AI systems had surpassed “all but the most skilled humans at finding and exploiting software vulnerabilities,” warning that the consequences for economies, public safety, and national security “could be severe.”

The gathering took place while bank executives were already in Washington for an industry event, with invitations largely extended to leaders of systemically important institutions. Regulators consider these banks critical to financial stability, meaning disruptions to their operations could have far-reaching consequences.

Attendees reportedly included David Solomon of Goldman Sachs, Brian Moynihan of Bank of America, Jane Fraser of Citigroup, Ted Pick of Morgan Stanley, and Charlie Scharf of Wells Fargo. Jamie Dimon of JPMorgan Chase was invited but did not attend.

In his annual shareholder letter released this week, Dimon cautioned that cybersecurity “remains one of our biggest risks,” adding that artificial intelligence “will almost surely make this risk worse.”

Anthropic said its yet-to-be-released Mythos model has already identified thousands of vulnerabilities across software and widely used applications. As a result, access to the system has been limited to a small group of companies, including Amazon, Apple, and Microsoft.

The move marks the first time the company has restricted a product rollout. Select infrastructure and technology groups, such as Cisco and Broadcom, have also been granted access, along with the Linux Foundation.

Concerns over misuse and undiscovered vulnerabilities

The developments come as fears grow that malicious actors could use advanced AI tools to uncover passwords or break encryption systems designed to protect sensitive data.

Anthropic said some of the flaws identified by Mythos date back as far as 27 years and had not been detected by developers or security monitors before the AI system surfaced them.

The Treasury meeting also follows a recent decision by the US government to classify Anthropic as a potential supply chain risk, a designation the company is currently challenging in court.

Despite the ongoing regulatory scrutiny and a supply chain risk designation from the U.S. Department of Defense, Anthropic has reported unprecedented financial momentum.

In a recent blog post released on April 6, the company said its annualized revenue run rate exceeded $30 billion as of early April 2026, more than tripling from roughly $9 billion at the end of 2025

Part of that growth has been driven by new compute partnerships with Google and Broadcom, highlighting rising demand for large-scale AI infrastructure. This agreement secures multiple gigawatts of next-generation TPU capacity to power frontier Claude models through 2027 and beyond

Its agentic coding platform, Claude Code, has emerged as a key contributor, generating more than $2.5 billion in run-rate revenue as of February.

Weekly active users on the platform have also doubled since the start of the year, pointing to rapid adoption of AI-driven development tools as the company shifts its focus toward high-value enterprise agents.