The Federal Reserve spent months without access to Anthropic's advanced AI model despite warning the nation's top banks it posed unprecedented cybersecurity threats.

In April, the Fed and Treasury Department convened an extraordinary meeting with bank CEOs to sound the alarm about Claude Mythos Preview. The model excels at identifying software vulnerabilities and was released to select institutions through Project Glasswing, Anthropic's cybersecurity initiative.

Yet the Fed itself lacked access to Mythos for at least three months afterward, leaving the central bank vulnerable while other institutions began patching their weaknesses.

Fed Chairman Kevin Warsh confirmed in Senate testimony last week that the central bank was still seeking access to Mythos as of July 15. "We are not the deciders as to who has access, but I have not been shy in sharing my views," Warsh told Senator Jack Reed.

The chairman said the Fed needed access "not just for the Federal Reserve but for other institutions to a whole range of these new artificial intelligence models so that they can protect themselves."

Anthropic initially granted access to roughly 50 organizations in April, naming JPMorgan Chase, Amazon, Apple, and Google among participants. The company expanded Project Glasswing in June to more than 150 organizations across 15 countries.

Regulatory Chaos Complicates Rollout

Mythos faced significant disruptions under the Trump administration's tighter AI oversight. In June, Anthropic disabled access to updated versions Mythos 5 and Fable 5 to comply with export control directives citing national security concerns.

Commerce Secretary Howard Lutnick later granted permission to restore access for "trusted partners," and the export controls were eventually lifted entirely.

The scrambling reflects broader chaos in AI policy leadership. Chris Fall, head of the Center for AI Standards and Innovation, resigned just three months after his appointment. Former White House AI czar David Sacks stepped down in March.

Daniel Newman, CEO of research firm The Futurum Group, expressed surprise at the Fed's exclusion. "You would think that the financial institution that drives policy for the rest would be front and center," he said.

It remains unclear whether the Fed has since gained access to Mythos or when vulnerability patching work might begin.