Mozilla has announced that Anthropic's latest AI model, Claude Mythos, identified 271 security vulnerabilities in the Firefox browser during internal testing. This marks a significant increase from a previous version of the model, which detected only 22 issues. All vulnerabilities have been patched this week. Mozilla noted that the vulnerabilities were within the scope of what top human researchers could find. Claude Mythos, released in March 2026, is Anthropic's most advanced model for reasoning, coding, and cybersecurity, and is available through the "Project Glasswing" initiative to select partners like Amazon, Apple, and Microsoft.