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

Mozilla engineers announced that they used Anthropic's Mythos AI to identify 271 security vulnerabilities in the Firefox browser over a two-month period, with what they described as 'almost no false positives.' The findings were presented by Mozilla's security team, who stated that the AI tool was able to detect flaws that had previously gone unnoticed.

Mythos is a large language model developed by Anthropic. Mozilla's team reported that the false positive rate was extremely low, meaning nearly every vulnerability flagged by the AI was a genuine security issue. The 271 vulnerabilities were discovered across various components of Firefox, though specific details of the flaws have not been publicly disclosed.

The use of AI for vulnerability discovery is a growing trend in the software industry. Mozilla's success with Mythos suggests that AI could become a standard tool for browser security testing, potentially reducing the need for manual code review. Mozilla has not announced whether it plans to adopt Mythos permanently or expand its use to other products.