U.S. financial regulators have issued a quiet but urgent warning to major banks after concerns emerged that next-generation artificial intelligence systems could significantly increase the sophistication and scale of cyberattacks targeting financial institutions.
Senior Wall Street executives were summoned to an unscheduled meeting in Washington by U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell. The discussion focused on the potential cybersecurity risks posed by Anthropic’s latest AI model, Mythos, which regulators believe could dramatically change the cyber threat landscape.
The meeting took place at the U.S. Treasury headquarters and included leaders from the country’s largest banks, many of whom were already in the city for a gathering of the Financial Services Forum, a lobbying group representing major U.S. financial institutions.
Growing concern over AI-enabled cyber threats
Officials used the meeting to warn banks about the emerging risks associated with advanced AI models capable of identifying and exploiting software vulnerabilities. According to regulators, models such as Anthropic’s Mythos could potentially enable attackers to automate complex cyber intrusions across widely used operating systems and web browsers.
While financial institutions already face persistent cyber threats, regulators are increasingly concerned that AI could significantly accelerate both the discovery of security weaknesses and the execution of attacks.
The banks invited to the meeting are all designated as systemically important financial institutions (SIFIs), meaning their operational stability is considered critical to the functioning of the global financial system. Any disruption to their systems could have widespread economic consequences.
Systemic risk concerns draw Federal Reserve attention
Jerome Powell’s participation highlighted the seriousness of the issue. The Federal Reserve plays a central role in supervising large banks and maintains deep visibility into their operational resilience through its examination network.
Officials stressed that the objective of the meeting was to ensure banks are proactively strengthening their cybersecurity defences and preparing for a new generation of AI-driven threats.
The urgency surrounding the issue was further underscored by comments from National Economic Council Director Kevin Hassett, who indicated that government agencies are closely reviewing the potential risks before allowing broader public access to the technology.
According to Hassett, Anthropic has agreed to delay wider public release of the Mythos model while officials assess its security implications.
Controlled rollout of powerful AI tools
Anthropic has acknowledged the potential risks associated with highly capable AI systems that can identify vulnerabilities in digital infrastructure. As a precaution, the company has so far limited access to Mythos, providing it only to a small group of major technology firms and financial institutions.
Regulators are now working with banks and technology companies to better understand how such systems could be misused by malicious actors and how financial institutions should adapt their defensive strategies.
What this means for the industry
- Advanced AI models may dramatically increase the speed and sophistication of cyberattacks against financial institutions.
- Regulators are treating AI-driven cyber threats as a potential systemic risk to the global financial system.
- Large banks are being urged to strengthen cyber defences and prepare for AI-assisted attack methods.
- AI developers may face increasing regulatory oversight before releasing highly capable models publicly.
- Collaboration between regulators, banks, and AI firms is becoming critical to managing emerging technology risks.
Photo by Boitumelo
