The Federal Reserve, alongside the U.S. Treasury, convened a special emergency meeting in April with the CEOs of the nation's leading banks. Officials issued a warning: a new advanced artificial intelligence model could pose an unprecedented cybersecurity threat to America's top financial institutions.
Developed by Anthropic, this model, named the ClaudeMythos preview, is adept at uncovering software vulnerabilities and security flaws. Through a cybersecurity initiative called Project Glasswing, Anthropic granted access to this model to a select group of banks and institutions.
However, for at least three months following this, the Federal Reserve itself was unable to obtain access to the Mythos model. While financial institutions were using the tool to patch their security systems, the Fed, as the world's most systemically important financial regulator, found itself potentially exposed to the very risks it was warning about.
As of July 15th, the Federal Reserve was still in the process of applying for access to the Mythos model, and it remains unclear whether the central bank has since successfully gained entry.
Previous reports indicated that the April meeting was led by then-Fed Chair Jerome Powell, with Treasury Secretary Scott Bessent also present to meet with banking executives. In a less prominent congressional hearing last week, Powell's successor, current Fed Chair Kevin Warsh, acknowledged to the Senate that he was still actively seeking to secure access to the Mythos model and other cutting-edge AI tools.
Responding to inquiries from Senator Jack Reed of Rhode Island regarding Mythos, Warsh stated, "Access to the model is not within our control. However, I have clearly communicated the existing security vulnerabilities to various federal agencies, advocating not just for the Fed but for all public institutions to gain access to a full suite of advanced AI models, enabling everyone to conduct necessary system protection work."
Warsh further clarified that the Federal Reserve's need extends beyond just the Mythos model.
"I don't want to focus solely on the Mythos model," Warsh said. "As the application of these advanced AI models continues to expand, the entire banking system, including the Federal Reserve, must do everything possible to patch all existing security vulnerabilities."
Warsh, who has served as Fed Chair for a relatively short time, is a strong proponent of artificial intelligence, describing it as a transformative technology.
Without access to the Mythos model, the Federal Reserve's efforts to identify vulnerabilities and strengthen its own security defenses would be significantly hampered.
Comments