Anthropic Unveils Powerful Mythos AI for Cybersecurity

Anthropic has previewed its new frontier AI model Mythos through Project Glasswing, a cybersecurity initiative involving 12 partners including Amazon, Apple, Microsoft, and Cisco. The model, described as Anthropic's most powerful yet, will scan software for vulnerabilities and has already identified thousands of critical zero-day flaws, some decades old. Mythos was previously leaked as 'Capybara' after draft documents were left in an unsecured public data cache. The preview will not be broadly released, though 40 organizations total will gain access. Anthropic says it has held discussions with federal officials, despite an ongoing legal dispute with the Trump administration over Pentagon-imposed supply-chain risk designations.