CoinTelegraph reports — The Federal Reserve cannot access an advanced AI model called "Claude Mythos Preview" for at least three months. Following a special meeting between the Federal Reserve and the Treasury Department, banks were warned that the model may pose significant cybersecurity risks. Other institutions are using this time to patch their own vulnerabilities.
According to Huitong Finance APP: In April, the Federal Reserve and the U.S. Treasury jointly convened a special meeting with the CEOs of leading U.S. banks to warn of a significant cybersecurity threat posed by a new AI model capable of precisely identifying vulnerabilities in financial institutions’ systems. However, as of July 15, the Federal Reserve had still not gained access to the model, even as other institutions used it to patch their vulnerabilities, leaving the Federal Reserve—the core of the global financial system—exposed to risk for an extended period.
Amid the confusion of the Trump administration's AI regulatory policies and the rapid rise of AI models from Asian major powers, the Federal Reserve has been forced into a reactive stance on financial security, putting the U.S. financial system's AI defense line under severe scrutiny.
The AI model named Claude Mythos Preview, developed by Anthropic, excels at precisely identifying software weaknesses and security vulnerabilities and has been integrated into the Project Glasswing cybersecurity initiative, made available on a selective basis to 50 curated institutions, including JPMorgan Chase, Amazon, and Apple. However, since at least three months after the April meeting, the Federal Reserve has still not been granted access to this model, leaving it unable to assess risks or identify vulnerabilities while other institutions actively patch security flaws—making the Fed a security weak point within the financial system.
Then-Fed Chair Jerome Powell led a special meeting in April, with Treasury Secretary Scott Bessent also participating, underscoring the official concern over the AI model’s security risks. However, as of July 15, the Fed was still actively seeking access, and there was no clear update on whether it would succeed.
Ansap did not respond to requests for comment, and the Federal Reserve declined to comment on the matter; the specific reason for the lack of access remains a mystery.
Kevin Warsh, the successor to Powell and Chair of the Federal Reserve, disclosed during last week’s congressional hearing that efforts are still underway to obtain access to Myos and other cutting-edge AI models. Warsh stated that access to these models is not determined by the Federal Reserve, but he has repeatedly highlighted associated security risks to various federal agencies, advocating not only for the Fed’s own access but also for enabling other institutions to secure model access to enhance their self-protection. He emphasized that the focus is not limited to Myos alone; as AI models become more widespread, the Federal Reserve and the U.S. banking system must urgently address and patch their own security vulnerabilities.
Daniel Newman, CEO of Futurum Group, expressed surprise, noting that as the central authority overseeing U.S. financial policy, the Federal Reserve should have been granted priority access to evaluate this new technology. The current lack of access highlights the lag between U.S. financial regulation and emerging technologies. Industry experts generally agree that without access to this model, the Federal Reserve cannot fully identify vulnerabilities in its own systems or effectively implement corresponding security measures.
The expansion of the Myos model has also been hampered by Ansapo’s complex relationship with the Trump administration. In June, under federal export control directives, Ansapo suspended access to the Myos 5 upgrade and Feibo 5 models; access was later partially restored only for “trusted partners” after approval by Commerce Secretary Howard Lutnick. Although subsequent restrictions were fully lifted, they exposed the chaos in U.S. AI regulation. Additionally, frequent personnel changes within the Trump administration’s AI regulatory team further intensified policy uncertainty, as Chris Fall, head of the Center for AI Standards and Innovation, resigned after only three months in office.
On the external competitive front, major Asian countries have made rapid progress in open-source large models; Moonshot AI’s Kimi K3 model has outperformed U.S. counterparts in multiple industry benchmark tests, raising concerns among American tech circles and government officials. Venture capitalist David Sacks bluntly stated that America’s self-imposed constraints in AI could cause it to lose the global AI race. Newman noted that the Federal Reserve, unable to access cutting-edge models, will be forced into a reactive stance, and given the extremely fast pace of technological iteration, even minor delays could introduce new security risks.
Overall, the Federal Reserve has been placed in a reactive position regarding financial security due to its lack of access to cutting-edge AI models—a situation stemming not only from limitations in model openness but also from the disarray and lag in U.S. AI regulatory policies. With cybersecurity vulnerabilities in the U.S. financial system still unaddressed and global AI competition intensifying, the Federal Reserve risks undermining its own regulatory effectiveness and potentially jeopardizing the stability of the entire global financial system if it fails to quickly close its technological gap. Moving forward, it must accelerate collaboration with relevant companies and agencies to overcome access restrictions at the earliest opportunity and strengthen its financial security defenses.
Amid the confusion of the Trump administration's AI regulatory policies and the rapid rise of AI models from Asian major powers, the Federal Reserve has been forced into a reactive stance on financial security, putting the U.S. financial system's AI defense line under severe scrutiny.
AI models conceal security threats, with the Fed lacking long-term options
The AI model named Claude Mythos Preview, developed by Anthropic, excels at precisely identifying software weaknesses and security vulnerabilities and has been integrated into the Project Glasswing cybersecurity initiative, made available on a selective basis to 50 curated institutions, including JPMorgan Chase, Amazon, and Apple. However, since at least three months after the April meeting, the Federal Reserve has still not been granted access to this model, leaving it unable to assess risks or identify vulnerabilities while other institutions actively patch security flaws—making the Fed a security weak point within the financial system.
Then-Fed Chair Jerome Powell led a special meeting in April, with Treasury Secretary Scott Bessent also participating, underscoring the official concern over the AI model’s security risks. However, as of July 15, the Fed was still actively seeking access, and there was no clear update on whether it would succeed.
Ansap did not respond to requests for comment, and the Federal Reserve declined to comment on the matter; the specific reason for the lack of access remains a mystery.
The Federal Reserve Chair is actively pushing forward, while various parties voice concerns about regulatory policies.
Kevin Warsh, the successor to Powell and Chair of the Federal Reserve, disclosed during last week’s congressional hearing that efforts are still underway to obtain access to Myos and other cutting-edge AI models. Warsh stated that access to these models is not determined by the Federal Reserve, but he has repeatedly highlighted associated security risks to various federal agencies, advocating not only for the Fed’s own access but also for enabling other institutions to secure model access to enhance their self-protection. He emphasized that the focus is not limited to Myos alone; as AI models become more widespread, the Federal Reserve and the U.S. banking system must urgently address and patch their own security vulnerabilities.
Daniel Newman, CEO of Futurum Group, expressed surprise, noting that as the central authority overseeing U.S. financial policy, the Federal Reserve should have been granted priority access to evaluate this new technology. The current lack of access highlights the lag between U.S. financial regulation and emerging technologies. Industry experts generally agree that without access to this model, the Federal Reserve cannot fully identify vulnerabilities in its own systems or effectively implement corresponding security measures.
Confusion in regulatory policies, combined with external competition, has significantly increased pressure on the Federal Reserve.
The expansion of the Myos model has also been hampered by Ansapo’s complex relationship with the Trump administration. In June, under federal export control directives, Ansapo suspended access to the Myos 5 upgrade and Feibo 5 models; access was later partially restored only for “trusted partners” after approval by Commerce Secretary Howard Lutnick. Although subsequent restrictions were fully lifted, they exposed the chaos in U.S. AI regulation. Additionally, frequent personnel changes within the Trump administration’s AI regulatory team further intensified policy uncertainty, as Chris Fall, head of the Center for AI Standards and Innovation, resigned after only three months in office.
On the external competitive front, major Asian countries have made rapid progress in open-source large models; Moonshot AI’s Kimi K3 model has outperformed U.S. counterparts in multiple industry benchmark tests, raising concerns among American tech circles and government officials. Venture capitalist David Sacks bluntly stated that America’s self-imposed constraints in AI could cause it to lose the global AI race. Newman noted that the Federal Reserve, unable to access cutting-edge models, will be forced into a reactive stance, and given the extremely fast pace of technological iteration, even minor delays could introduce new security risks.
Summary
Overall, the Federal Reserve has been placed in a reactive position regarding financial security due to its lack of access to cutting-edge AI models—a situation stemming not only from limitations in model openness but also from the disarray and lag in U.S. AI regulatory policies. With cybersecurity vulnerabilities in the U.S. financial system still unaddressed and global AI competition intensifying, the Federal Reserve risks undermining its own regulatory effectiveness and potentially jeopardizing the stability of the entire global financial system if it fails to quickly close its technological gap. Moving forward, it must accelerate collaboration with relevant companies and agencies to overcome access restrictions at the earliest opportunity and strengthen its financial security defenses.