Odaily Planet Daily reports that Anthropic’s latest AI model, Claude Mythos, identified 271 security vulnerabilities during testing in Mozilla’s Firefox browser; the issues have recently been resolved, highlighting AI’s potential in cybersecurity.
Mozilla states that this model can efficiently analyze large codebases and identify potential vulnerabilities—a process that previously relied heavily on scarce senior security researchers. In earlier tests, another Claude model discovered 22 critical vulnerabilities.
Nevertheless, Mozilla believes that the vulnerabilities discovered by AI are still within the capabilities of top human researchers, but their scalability could shift the balance of offense and defense, giving defenders a systemic advantage for the first time. However, industry experts have warned that if such technologies are misused, they could exacerbate the risk of automated attacks. (Decrypt)
