Anthropic Accidentally Leaks 'Claude Mythos' Model Details — Cybersecurity Stocks Plunge, Internal Docs Cite 'Unprecedented Risks'

Anthropic accidentally exposed details about its unreleased AI model named 'Claude Mythos' on its website, as discovered and reported by Fortune. The leaked data included draft blog posts, images, and PDFs — nearly 3,000 assets that had not been previously made public.

Most notably, internal documents revealed that Anthropic considers Mythos to be its most capable model to date, while also acknowledging that it poses 'unprecedented cybersecurity risks.' The documents also referenced a new model tier called 'Capybara,' described as larger and more capable than Anthropic's current top-tier Opus model.

Following the leak, cybersecurity stocks experienced a significant selloff. Major companies including CrowdStrike (CRWD) and Palo Alto Networks (PANW) saw sharp declines as investors grew concerned about the potential for advanced AI models to be weaponized for sophisticated cyberattacks. The market reaction underscores the financial community's acute awareness of AI's dual-use nature.

The incident comes at a particularly sensitive time for Anthropic, which is reportedly considering an IPO in October 2026 with a projected valuation exceeding $60 billion. The security lapse raises questions about the company's internal controls and highlights the critical importance of information security for companies developing cutting-edge AI technology.