Anthropic has released a preview of its new frontier AI model, Mythos, calling it one of its most powerful yet. It will be used by 12 partner organizations for defensive cybersecurity work under Project Glasswing. While not specifically trained for security, Mythos will scan software for code vulnerabilities. Anthropic says it has already identified thousands of critical zero-day flaws, some decades old. Partners include Amazon, Apple, Cisco, Microsoft, and the Linux Foundation. The model’s existence was previously leaked in a data security incident, where it was referred to internally as “Capybara.” Anthropic is also in discussions with federal officials, despite an ongoing legal battle with the Trump administration over defense contracting rules.
