Federal Reserve Chairman Kevin Warsh testifies during a Senate Banking, Housing, and Urban Affairs Committee hearing entitled “Semiannual Monetary Policy Report to Congress” on Wednesday, July 15, 2026, in the Dirksen Building.
Tom Williams | Cq-roll Call Inc. | Getty Images
In April, the Federal Reserve and the Treasury Department convened a special meeting with the CEOs of the nation’s largest banks. Officials have sounded the alarm about advanced new artificial intelligence models that could pose an unprecedented cybersecurity threat to the nation’s top financial institutions.
Anthropic, which developed the AI model Claude Mythos Preview, said the product excels at identifying weaknesses and security vulnerabilities within software. The company released it to a select group of banks and other institutions as part of a cybersecurity initiative called Project Glasswing.
The Fed itself was unable to access Mythos for at least three months afterward, leaving perhaps the most systemically important global financial institution vulnerable even as other institutions began to fill in the gaps.
The Feds were still trying to access Mythos as of July 15th. It is not clear whether the central bank has since gained access to the model.
Anthropic did not respond to a request for comment. The Federal Reserve declined to comment for this article.
As CNBC previously reported, the April meeting was held under former Federal Reserve Chairman Jerome Powell, along with Treasury Secretary Scott Bessent and bank CEOs. But in Congressional testimony last week that received little attention, Mr. Powell’s successor, Speaker Kevin Warsh, told the Senate that he continues to work to secure access to Mythos and other cutting-edge AI models.
“While I don’t have the power to decide who has access, I have unashamedly shared my views on vulnerabilities with authorities across the government and have asked not just the Federal Reserve, but other agencies, for access to all of these new artificial intelligence models so they can protect themselves,” Warsh said in response to a question about Mythos.
Warsh revealed there were other models the Fed needed access to as well.
“However, we don’t want to just isolate Mythos,” Warsh said. “As these new models become more widespread, our banking system, and frankly the Federal Reserve, needs to do everything in its power to fix whatever vulnerabilities we have.”
Warsh embraced AI during his short time at the Fed, calling it a transformative technology.
It is not clear whether work can begin to address vulnerabilities without accessing Mythos.
Anthropic announced Claude Mythos Preview and Project Glasswing in early April.
The company said about 50 organizations had access to the model at the time, but named only a handful, including banks. JP Morgan Chaseand tech giants such as Amazon, apple and google. Anthropic also said it is in “ongoing discussions with U.S. government officials” about the model, including the Cybersecurity and Infrastructure Security Agency and the Center for AI Standards and Innovation.
Anthropic expanded access to Project Glasswing in June, adding more than 150 organizations in 15 countries to the initiative. Daniel Newman, CEO of research firm Futurum Group, said he was surprised to hear the Fed was not included.
“The kind of financial institutions that drive all the policies of the rest of the financial institutions will be at the center of the opportunity to at least evaluate new technology,” he told CNBC on Tuesday.
The development of Mythos has sparked both intrigue and confusion in recent months, especially as Anthropic’s rocky relationship with the Trump administration has caused complications.
Anthropic announced in June that it needed to disable access to the Mythos 5, an updated version of the model, and the Fable 5, the more widely released version of the Mythos, to comply with an export control directive from the federal government that referred to “national security authorities.”
Commerce Secretary Howard Lutnick has given the company permission to restore access to Mythos to a select group of “trusted partners,” according to a letter seen by CNBC. Export restrictions were subsequently completely lifted.
Since President Donald Trump signed the AI Executive Order in June, the Trump administration has taken a more active role in regulating AI. But questions remain about who is deciding AI policy behind the scenes, and the turmoil at the Fed, where President Trump chose Warsh to lead, is another sign of confusion.
Chris Fall, director of the Center for AI Standards and Innovation, has stepped down as director just three months after being selected for the position by the Trump administration, CNBC confirmed Monday. Venture capitalist David Sachs previously held the position of White House AI and cryptocurrency czar, but resigned from that role in March.
Pressure is mounting.
China’s indiscriminate model is gaining momentum against major products from U.S. companies such as OpenAI and Anthropic, raising concerns from technology executives and government officials about whether the U.S. lead in the AI race will persist. Earlier this month, Chinese startup Moonshot AI released a model called Kimi K3 that outperforms these companies on some industry benchmarks.
In a post on X on Friday, Sachs said Kimi K3’s performance was “concerning” and that “America is pushing itself into a corner.”
“This is how we lose the AI race,” he wrote. “If we get stuck, other countries won’t follow our rules either.”
Futurum’s Mr. Newman said the Fed lacks access to the most advanced models that other agencies have, so it’s “definitely going to play catch-up.”
“Every day and week, technology leaders within large organizations are faced with a constant barrage of online innovations, whether it’s Chinese innovation or American innovation,” he said.
