Anthropic Claude Mythos Discovers 271 Vulnerabilities in Firefox Browser

DeepTech TechFlow reports that on April 22, according to Decrypt, Mozilla recently disclosed that Anthropic's latest AI model, Claude Mythos, identified 271 security vulnerabilities during internal testing of the Firefox browser. The related vulnerabilities have been fixed this week. In comparison, a previous version of Anthropic's model had only discovered 22 security-sensitive vulnerabilities. Mozilla stated that all identified vulnerabilities were within the scope of what top human researchers could uncover. Claude Mythos was officially released in March 2026 and is Anthropic's most powerful model in reasoning, coding, and cybersecurity. It is currently available exclusively to vetted partners such as Amazon, Apple, and Microsoft through the "Project Glasswing" initiative.