The Federal Reserve is currently struggling to secure access to advanced artificial intelligence models, including Anthropic’s “Claude Mythos Preview,” which officials believe are essential for protecting the nation’s financial infrastructure. Despite warnings from the Treasury Department and the Fed about AI-driven cybersecurity risks, Chairman Kevin Warsh confirmed in mid-July 2026 testimony that the central bank remains in a reactive position, lacking the direct access already granted to private financial institutions.
Federal Reserve Access Gaps and Financial Security
In April, the Federal Reserve and the Treasury Department convened a meeting with CEOs from top U.S. banks to address the cybersecurity threats posed by emerging AI. Anthropic’s Claude Mythos Preview, a core component of the “Project Glasswing” initiative, was identified as a tool capable of pinpointing software vulnerabilities. While private entities such as JPMorgan Chase, Amazon, Apple, and Google were granted access to the model, the Federal Reserve itself lacked the same capability for at least three months following the initial warnings.
During a Senate Banking, Housing and Urban Affairs Committee hearing on July 15, 2026, Chairman Kevin Warsh acknowledged the disparity. “I have not been shy in sharing my views with authorities across the government about the vulnerabilities,” Warsh stated in response to Sen. Jack Reed, D-R.I. Warsh emphasized that the Fed’s lack of access is not merely a matter of missing one tool, but a broader need to integrate various cutting-edge models to defend the banking system against evolving threats.
Did you know?
Project Glasswing, launched by Anthropic in April, expanded its reach significantly by June, granting access to more than 150 organizations across 15 countries, yet the Federal Reserve remained outside this initial deployment loop.
Regulatory Friction and Internal Turnover
The path to AI integration at the Federal Reserve has been complicated by shifting regulatory oversight within the Trump administration. Anthropic’s rollout of Mythos faced direct interference in June when the company was forced to disable access to updated versions of the model to comply with federal export controls citing “national security authorities.”
While Commerce Secretary Howard Lutnick eventually granted permission for select partners to regain access and the controls were later lifted, the episode highlighted the volatility of current AI policy. This turbulence is compounded by significant turnover in key government AI roles. Chris Fall, the director of the Center for AI Standards and Innovation, resigned in July, just three months into his tenure. His departure followed the exit of White House AI and crypto czar David Sacks in March, leaving a vacuum in leadership as the Fed attempts to navigate these new technological mandates.
Global Competition and the AI Arms Race
The delay in domestic AI adoption is occurring as international competitors rapidly advance. Chinese startups are closing the technological gap, with the recent release of Moonshot AI’s “Kimi K3” model, which has reportedly outperformed U.S.-based models on specific industry benchmarks. Daniel Newman, CEO of the Futurum Group, noted that the Fed is now in a position where it must “play catch up.”
According to Newman, the speed of innovation—both domestic and international—creates a constant pressure on large financial institutions. Sacks, the former White House AI official, expressed concern on social media, suggesting that domestic regulatory hurdles are causing the U.S. to “tie itself in knots,” potentially ceding its long-standing lead in the global AI race.
Frequently Asked Questions
- Why does the Federal Reserve need access to Claude Mythos?
Officials view the model as a critical tool for identifying and patching software vulnerabilities within the nation’s financial systems. - Was the Federal Reserve denied access to AI models?
Chairman Kevin Warsh stated the Fed is actively working to secure access, noting that the central bank is not the final decision-maker regarding who receives the technology. - What is Project Glasswing?
It is an initiative by Anthropic that provides select organizations with access to AI models designed to detect cybersecurity weaknesses.
Stay informed on the intersection of finance and technology. Subscribe to our daily newsletter for the latest updates on central bank policy and AI regulation.
Worth a look