Bessent Warns of Threat of AI-Powered Bank Account Hacks

• Bessent and Fed Chair Jerome Powell met with major banks like JPMorgan and Bank of America, directing them to use the Mythos AI model to find vulnerabilities in their systems
• Anthropic's Mythos has discovered thousands of high-severity security flaws, raising concerns about AI systems that can execute or orchestrate multistep cyberattacks at scale
• Officials emphasized the need to balance safety and innovation as AI capabilities advance, with cybersecurity risk evolving from targeted attacks to 'ambient exposure' where organizations face continuous automated scanning and probing

Summary: Bessent Warns of AI-Powered Bank Account Hacks

Treasury Secretary Bessent has issued a stark warning about AI-powered cyberthreats targeting American banks, confirming concerns following a May 3, 2026 meeting with Federal Reserve Chair Jerome Powell and Wall Street executives.

Key Development:

The meeting focused on Anthropic's "Mythos" AI model, which has discovered thousands of high-severity vulnerabilities in major operating systems and web browsers. Officials instructed major banks to use this model proactively to identify defensive weaknesses in their systems.

Critical Statement:

When asked directly if Americans should worry about AI hacking their bank accounts, Bessent responded bluntly: "You should." He emphasized that recent advancements represent a "step change" in large language model capabilities, with more expected from competing AI companies.

Market Implications:

The threat is double-edged—while banks, payment processors, and infrastructure providers can use AI tools to identify and patch vulnerabilities, the same capabilities enable hackers to accelerate discovery and exploitation of systemic flaws across the financial ecosystem.

Security Landscape Shift:

Cybersecurity experts note that AI has evolved beyond being merely a tool for attackers; it now "replicates aspects of the attacker itself." The risk has transformed from targeted attacks to "ambient exposure," where organizations face continuous scanning and probing by AI systems operating at scale.

Policy Balancing Act:

Bessent stressed the need for "very important calculus" between safety and innovation to maintain U.S. competitiveness, though no specific protective measures were disclosed.

The warning highlights particular vulnerability for enterprises with inconsistent patching, over-permissioned accounts, and weak configuration management facing AI-orchestrated multi-step intrusion attempts.