The U.S. Treasury and the Federal Reserve convened an emergency meeting with top Wall Street bank executives to assess emerging cyber risks posed by advanced artificial intelligence, according to Bloomberg.
On April 7, U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned chief executives from major financial institutions to a closed-door meeting at the Treasury Department in Washington, United States.
Attendees included CEOs from systemically important financial institutions such as Citigroup, Morgan Stanley, Bank of America, Wells Fargo and Goldman Sachs. Jamie Dimon, CEO of JPMorgan Chase, did not attend, Bloomberg reported, citing sources.
The meeting was convened to discuss the potential impact of “Mythos,” a new AI model developed by Anthropic, on the financial system.
Mythos is believed to go beyond identifying vulnerabilities in major operating systems and web browsers, with the capability to generate code that could exploit those weaknesses in real-world attacks.
Unlike earlier AI tools primarily used for coding assistance or analysis, Mythos combines vulnerability detection with execution, raising concerns that it could accelerate the automation of cyberattacks. Experts warn the model may signal a shift from AI-assisted hacking to fully automated cyber intrusions.
Particular concern centers on its ability to identify “zero-day” vulnerabilities—security flaws unknown even to developers—potentially exposing critical systems before patches are available. Regulators view the model as a potential threat to core infrastructure, including the financial sector, Bloomberg reported.
The meeting, organized on short notice without prior public disclosure, reflects how authorities are treating the issue as a systemic risk rather than a purely technological concern.
Anthropic has limited access to Mythos due to its potential risks, making the model available only to select technology and financial firms. The company is also working with more than 40 infrastructure organizations, including Nvidia, Amazon, Apple and JPMorgan Chase, through an initiative known as Project Glasswing to evaluate defensive systems.
Bloomberg reported that Anthropic had also engaged with the U.S. government prior to the model’s release to discuss its offensive and defensive cyber capabilities.