In a recent development, the Federal Reserve has been grappling with limited access to an advanced AI model, Claude Mythos Preview, for over three months. This situation arises after an urgent meeting in April with U.S. bank CEOs, where the Fed warned about the cybersecurity risks the model could pose to major financial entities. However, without access, the Fed finds itself potentially vulnerable, even as other institutions, who did have access, began to address vulnerabilities.
Developed by the company Anthropic, the Claude Mythos Preview excels at identifying software weaknesses. The model has been rolled out to a select group of banks and institutions as part of a cybersecurity project named Project Glasswing. However, the Fed, despite its status as the principal financial authority, found itself without access, bringing forth concerns about safeguarding critical financial infrastructure.
While the Fed is still striving to secure access, the model remains unavailable to them. According to testimony before Congress by Kevin Warsh, the current Fed chairman, there is an ongoing attempt to secure not only Claude Mythos, but also other cutting-edge AI models to help protect the nation’s banking system. Warsh emphasized the need for the Fed to address vulnerabilities by adopting these new AI technologies, yet access remains elusive.
Anthropic, the creator of Mythos, and other involved parties, including the Fed, have refrained from commenting further on the matter. However, sources confirmed that the Mythos has been distributed to roughly 50 organizations, including major banks like JPMorgan Chase, and tech giants such as Amazon, Apple, and Google.
When reaching out to experts, many expressed surprise that such a crucial entity as the Fed was not among those with initial access. The concern is that without this advanced AI tool, the Fed might be at a disadvantage in responding to cybersecurity threats that other institutions have begun to address.
Amidst this landscape, geopolitical tensions and policy decisions further complicate the deployment of such technologies. In June, amid some restrictions, the Trump administration ordered compliance with export controls citing national security concerns, impacting access for some entities. However, this was later reversed, allowing the continued rollout to select partners.
As the situation unfolds, competition in AI technology heats up. Companies, including Chinese startups, are developing models that threaten to surpass those from U.S. firms. For experts, this scenario signifies a race where the U.S. could potentially fall behind due to internal regulatory challenges.
Moving forward, pressure mounts on the Fed to ‘catch up’ as other financial and tech institutions continue to explore AI-driven defenses against security vulnerabilities, an undertaking made significantly difficult without the Mythos model.
You can read the original article here: [CNBC – Fed Mythos AI Cybersecurity Banks Project Glasswing](https://www.cnbc.com/2026/07/21/fed-mythos-ai-cybersecurity-banks-project-glasswing.html)
