Mozilla revealed that its latest Firefox build benefited from Anthropic’s Claude Mythos Preview model. The AI helped the team spot and fix 271 security holes before release. The move follows Anthropic’s Project Glasswing, which claims to use AI to guard against cyberattacks and secure critical services in the browser ecosystem.

Mozilla’s post emphasized that Mythos surfaced vulnerabilities humans could also find, suggesting the AI isn’t yet superior to skilled security researchers. The foundation noted no category or complexity that the model uncovered which a human couldn’t, given enough time and resources. Still, the partnership signals growing confidence that AI tools can augment traditional testing efforts.

For Firefox users, the update arrives with an option to disable all generative‑AI features, a move that appeases those wary of embedded AI. The 271 patches address a range of vulnerabilities, from privilege escalation to data leakage, tightening the browser’s security posture. Mozilla’s experiment demonstrates that AI can streamline vulnerability discovery without replacing human oversight.

The collaboration also offers a real‑world test for Anthropic’s Project Glasswing, which has faced criticism over its effectiveness. By proving useful in a major open‑source project, Claude Mythos gains credibility among security teams. Ultimately, the partnership shows that AI can be a practical aid in hardening software, rather than a speculative promise for developers and safety.