Mozilla's Use of AI Model Mythos Identifies 271 Vulnerabilities in Firefox
Mozilla has announced that its use of the AI model Mythos has successfully identified 271 security vulnerabilities in Firefox with almost no false positives. This development marks a significant improvement in AI-assisted vulnerability detection, which has historically been plagued by inaccuracies. Mozilla's engineers attribute this success to advancements in AI models and the development of a custom 'harness' that guides the AI through specific tasks. This harness allows the AI to access the same tools and pipeline used by human developers, enhancing its ability to accurately identify vulnerabilities. 
