Mozilla says 271 vulnerabilities found by Mythos have "almost no false positives"

Mozilla's use of Anthropic Mythos, an AI model for identifying software vulnerabilities, has yielded 271 vulnerabilities with almost no false positives, according to the company. This development is significant, as it suggests that AI-assisted vulnerability detection can be an effective tool for defenders. The results are particularly notable given the skepticism surrounding AI's ability to detect vulnerabilities, with some critics arguing that such claims are often exaggerated or lack nuance. Mozilla's CTO has stated that AI-assisted vulnerability detection could spell the end of zero-days, giving defenders a decisive advantage. The company's experience with Anthropic Mythos appears to bear this out, with the AI model identifying a substantial number of vulnerabilities without generating a large number of false positives¹. This matters to security practitioners, as it highlights the potential for AI to help close patching windows and reduce exposure to zero-day attacks.