Mozilla: Anthropic's Mythos found 271 security vulnerabilities in Firefox 150

Mozilla's collaboration with Anthropic's AI, Mythos, unearthed 271 security vulnerabilities in Firefox 150—28% more than the previous record for automated bug sweeps. The tool's success hinges on a novel reinforcement learning model that simulates attacker strategies, and it even uncovered a zero-day exploit in the browser's JavaScript engine, a vulnerability that existed for 14 months undetected. This breakthrough could redefine how browsers handle bug bounties.