Federal Reserve alerts on Anthropic's Mythos AI but lacks access for months

Fed's Alarm Over Mythos AI

In April, the Federal Reserve and Treasury Department held a meeting with major bank CEOs, highlighting the risks posed by Anthropic's advanced AI model, Claude Mythos Preview. This model is designed to identify weaknesses and vulnerabilities in software, which raised concerns about cybersecurity threats to the banking sector.

Project Glasswing Initiative

Anthropic introduced Mythos as part of its cybersecurity initiative, Project Glasswing. This select release aimed at providing banks and other institutions with tools to better secure their systems against potential threats.

Access Delays for the Fed

Despite the urgency of the situation, the Federal Reserve was unable to access Mythos until several months later. This delay left the Fed, which oversees significant financial operations, exposed while other institutions began to patch vulnerabilities identified by Mythos. As of mid-July, it remained unclear if the Fed had gained access to the model.

Responses from Fed Leadership

Jerome Powell, the previous Fed leader, highlighted the issue in congressional testimony, stating he had been seeking access to the AI model despite the challenges. Kevin Warsh, Powell's successor, reiterated the need for access to Mythos and other advanced AI systems to improve cybersecurity measures across the banking system.

Implications for the Financial Sector

The inability of the Fed to utilize Mythos emphasizes potential gaps in cybersecurity preparedness among major financial institutions. The situation illustrates the critical need for timely access to advanced technologies like Mythos to safeguard the nation's financial stability.