Anthropic has confirmed the existence of Claude Mythos, a highly advanced, unreleased AI model that the company is currently withholding due to significant cybersecurity and safety concerns.

Key points

• Anthropic accidentally leaked details about Claude Mythos, a model described as significantly more powerful than the current Claude Opus 4.6.
• The company is delaying the public release of Mythos, citing unprecedented cyber capabilities that pose potential security risks.
• A federal judge recently issued a temporary ruling blocking the Department of Defense from labeling Anthropic a security risk.
• Under Secretary of War Emil Michael publicly criticized Anthropic following the leak, despite the company holding government clearance for classified materials.
• Reports suggest Anthropic is considering an initial public offering (IPO) later this year, potentially leveraging the model's capabilities to attract investors.

The leak highlights the ongoing tension between AI developers and government agencies regarding the balance of national security, corporate autonomy, and public safety. Furthermore, the revelation of a high-performance model serves as a significant signal to the market as Anthropic prepares for a potential public offering.