Fed sought access to Anthropic's Claude Mythos Preview as banks flagged AI cybersecurity risks

According to CNBC, the Federal Reserve was still trying to secure access to Anthropic's Claude Mythos Preview on July 15 after U.S. officials warned in April that the AI model could pose an unprecedented cybersecurity threat to major banks. Anthropic said the model, released through its Project Glasswing initiative, was designed to find software weaknesses and security vulnerabilities, and it later expanded access to more than 150 organizations across 15 countries. Fed Chairman Kevin Warsh told the Senate he was still pressing for access to Mythos and other advanced AI models so the central bank and other institutions could better protect themselves.