Anthropic's Mythos AI Identifies 271 Security Vulnerabilities in Firefox 150

Mozilla reported on Tuesday that early access to Anthropic's Mythos Preview model helped identify 271 security vulnerabilities in Firefox 150 before its release. Firefox CTO Bobby Holley expressed optimism that defenders now have a decisive advantage in cybersecurity battles. This announcement followed Anthropic's earlier decision to limit Mythos Preview to select industry partners, citing its exceptional vulnerability detection capabilities. The model's performance significantly outpaced Anthropic's Opus 4.6 model, which found only 22 security-sensitive bugs in Firefox 148 last month. Debate continues about whether Mythos represents a breakthrough in AI-aided hacking or simply incremental advancement in AI capabilities. Mozilla's blog post provided concrete data showing the model analyzed unreleased Firefox source code to pre-identify these vulnerabilities.