The Fed rang the alarm about Anthropic’s Mythos AI — but went without it

In April, the Federal Reserve and the Treasury Department held an extraordinary meeting with the CEOs of the nation’s largest banks. Authorities have sounded the alarm about an advanced new artificial intelligence model that could pose an unprecedented cybersecurity threat to the nation’s top financial institutions.

Claude Mythos Preview Anthropic, the company behind the AI ​​model, said this offering is excellent at identifying weaknesses and vulnerabilities in software. The company rolled it out to a select group of banks and other institutions as part of a cybersecurity initiative called Project Glasswing.

For at least the next three months, the Fed had no access to Mythos, leaving arguably the most systemically important global financial institution vulnerable even as other institutions began to patch its vulnerabilities.

The Fed was still trying to access Mythos as of July 15. It is unclear whether the central bank has since gained access to the model.

Anthropic did not immediately respond to a request for comment. The Fed declined to comment for this article.

The April meeting was chaired by former Fed leader Jerome Powell, who brought together bank CEOs as well as Treasury Secretary Scott Bessent, CNBC previously reported.. But in little-noticed congressional testimony last week, Powell’s successor, Chairman Kevin Warsh, told the Senate he was still working to secure access to Mythos and other cutting-edge artificial intelligence models.

Sen in response to questions about Mythos. “We’re not the ones who decide who has access, but I haven’t hesitated to share my views about the vulnerabilities with officials across the government and request access to all of these new AI models so they can protect themselves, not just for the Federal Reserve but for other agencies as well,” Warsh told Jack Reed, D.-RI.

Warsh explained that there are other models the Fed needs to tap into.

“I wouldn’t want to isolate Mythos, though,” Warsh said. “As these new models become more widely available, our banking system and frankly the Federal Reserve need to do what we can to fix the vulnerabilities that we have.”

Warsh has embraced the adoption of artificial intelligence in his short time at the Fed, calling it a transformational technology.

Without access to Mythos, it’s not clear that work to fix the vulnerabilities can begin.

Anthropic introduced Claude Mythos Preview and Project Glasswing in early April.

The company said at the time that about 50 organizations had access to the model but only named a few, including a bank. JPMorgan Chaseand technology giants Amazon, Apple And Google. Anthropic said it was in “ongoing discussions with U.S. government officials” about the model, including the Cybersecurity and Infrastructure Security Agency and the AI ​​Standards and Innovation Center.

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 on this list.

“You would think that the financial institution that drives all the policy for the rest of the financial institutions would be front and center to at least have a chance to evaluate new technology,” he told CNBC on Tuesday.

The launch of Mythos has sparked both intrigue and confusion in recent months, particularly as Anthropic’s troubled relationship with the Trump administration has led to complications.

In June, Anthropic said it was forced to disable access to Mythos 5, an updated version of the model, and Fable 5, a more widely released version of Mythos, to comply with a federal government export control directive that cited “national security authorities.”

Commerce Secretary Howard Lutnick gave the company permission to restore access to Mythos to a select group of “trusted partners,” according to a letter seen by CNBC. Export controls were later completely removed.

The Trump administration has been taking a much more active role in AI regulations since President Donald Trump signed an AI executive order in June. But questions remain about who decides AI policy behind the scenes, and the infighting at the Fed, where Trump chose Warsh to lead, is another sign of chaos.

Chris Fall, head of the Center for Artificial Intelligence Standards and Innovation, resigned as director just three months after taking office. selected CNBC confirmed Monday that the Trump administration has accepted the job. Venture capitalist David Sacks previously served as artificial intelligence and cryptocurrency czar at the White House, but left the post in March.

The pressure is increasing.

China’s open-heavy models are gaining ground and sparking opposition to leading offerings from US companies like OpenAI and Anthropic concerns of technology executives and government officials about the durability of U.S. leadership in the AI ​​race. Moonshot AI, a Chinese startup, launched a model called the Kimi K3 earlier this month, and it beat those companies on some industry benchmarks.

In a post on X on Friday, Sacks said Kimi K3’s performance was “concerning” and that “America is in knots.”

“This is how you lose the AI ​​race,” he wrote. “If we get ourselves bogged down, the rest of the world won’t play by our rules.”

Futurum’s Newman said the Fed will “definitely be playing catch-up” because it doesn’t have access to the cutting-edge models that other institutions have.

“Every day, every week, technology leaders inside large institutions face a constant barrage of net innovation, whether it’s China’s innovation or U.S. innovation,” he said.