How Anthropic’s Mythos has rewritten Firefox’s approach to cybersecurity

Mozilla reported that Anthropic’s Mythos model has dramatically improved automated vulnerability discovery for Firefox, surfacing hundreds of high-severity bugs (including sandbox vulnerabilities and a 15-year-old parsing flaw) and enabling a large increase in shipped fixes; the researchers use AI to generate candidate patches but rely on human engineers to produce deployable fixes, and they caution that the same techniques could be used by attackers.