ON 21 April 2026, Mozilla said that Mythos Preview had helped pre-identify 271 security vulnerabilities in Firefox 150, the week’s release of the browser. In its blog post, Mozilla noted that the vulnerabilities were detected by analysing unreleased source code, and that the results were substantial enough to prompt their CTO Bobby Holley to say defenders now have a better chance in the ongoing cyber battle.
By comparison, Anthropic’s Opus 4.6 model found 22 security-sensitive bugs when analysing Firefox 148 last month, underscoring the rapid gains in AI-aided vulnerability analysis. Holley suggested that Mythos could identify bugs more efficiently than traditional manual methods, potentially reducing the cost and effort required to find issues.
He also remarked that this kind of AI-assisted defence is likely to be relevant for open source projects, which often rely on volunteer maintenance and publicly accessible codebases. According to Mozilla, the broader implication is that AI tools like Mythos could shift the cybersecurity balance toward defenders as software teams increasingly engage with AI-driven analysis.