‹ Back

US Regulators Warn Banks on Anthropic AI Risks | News

 

U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened an urgent meeting with major bank executives to address cybersecurity risks linked to Anthropic’s latest AI model.

 

 

Emergency meeting centers on AI cyber risks

 

The closed-door meeting, held at the U.S. Treasury Department in Washington earlier this week, brought together leaders of the country’s largest financial institutions to discuss the potential systemic implications of advanced AI systems.

 

According to reports first published by Bloomberg, the discussion focused on risks associated with Anthropic’s newly introduced AI model, known as Claude Mythos Preview.

 

U.S. officials used the meeting to warn that the model’s capabilities could expose financial institutions to new forms of cyber threats, particularly through its ability to identify vulnerabilities in software systems.

 

The session was convened on short notice, highlighting the urgency attached to the issue by regulators overseeing financial stability and cybersecurity preparedness.

 

Major banks briefed on vulnerability exposure

 

Executives from several systemically important U.S. banks attended the meeting, including representatives from Bank of America, Citigroup, Goldman Sachs, Morgan Stanley, and Wells Fargo.

 

JPMorgan Chase Chief Executive Jamie Dimon was invited but did not attend, according to the same reports.

 

Officials emphasized the need for banks to assess their exposure to vulnerabilities that could be identified or exploited using advanced AI systems. The discussions aimed to ensure that financial institutions are taking appropriate defensive measures to protect critical infrastructure.

 

The meeting also reflected growing coordination between financial regulators and private-sector institutions as artificial intelligence capabilities expand into domains with direct implications for financial security.

 

Anthropic limits rollout of Mythos model

 

Anthropic released the Claude Mythos Preview model to a limited group of partners earlier in the week, including major technology companies such as Amazon, Apple, and Microsoft.

 

The restricted rollout marked a departure from broader public releases typical of earlier AI systems, reflecting internal and external concerns about the model’s potential misuse.

 

The company has described the model as a general-purpose system with capabilities extending beyond cybersecurity, but reports indicate that its ability to detect and potentially exploit software vulnerabilities has drawn particular scrutiny.

 

Anthropic acknowledged the model could uncover previously unknown weaknesses across major operating systems and web browsers.

 

Regulators highlight systemic implications

 

The meeting signals increasing concern among U.S. policymakers that rapidly advancing AI technologies could introduce new categories of systemic risk within the financial sector.

 

Officials sought to ensure that banks are prepared for scenarios in which AI tools could be used by malicious actors to identify and exploit weaknesses at scale, potentially affecting critical financial infrastructure.

 

The decision to convene top bank executives reflects the significance regulators are placing on AI-driven cybersecurity threats, particularly as financial institutions continue to integrate advanced technologies into core operations.

 

The developments mark one of the most direct interventions by U.S. financial authorities in response to a specific AI system, highlighting the intersection of emerging technology risks and financial system oversight.