Aug 4, 07:30 PM

Google’s Big Sleep AI Spots 20 Flaws in Open-Source Software

Google said its artificial-intelligence bug-hunting system, Big Sleep, has identified 20 previously unknown security flaws across a range of widely used open-source projects, including the multimedia library FFmpeg and the image-editing suite ImageMagick. The company disclosed the findings on Monday, calling them the first batch of vulnerabilities reported by the tool since it was developed by Google DeepMind in collaboration with the Project Zero security team. Big Sleep is powered by Google’s Gemini large language model and operates largely autonomously, though every report is reviewed by a human expert before disclosure, Vice President of Security Heather Adkins said. Royal Hansen, a vice president of engineering, described the results as evidence of “a new frontier in automated vulnerability discovery.” Details of the individual flaws are being withheld until patches are available, in line with industry practice. The announcement adds to a growing roster of AI-driven security tools staking claims in vulnerability research. Earlier the same day, start-up Gecko Security said its own AI scanner had uncovered more than 30 zero-day bugs in open-source projects. While proponents argue such systems can scale code review and cut false positives, software maintainers warn that AI-generated reports can also contain errors, underscoring the need for human validation.

#Google #Big Sleep #FFmpeg #ImageMagick #Google DeepMind #Project Zero #Gemini #Heather Adkins #Royal Hansen #Gecko Security

Written with ChatGPT .

Google’s Big Sleep AI Spots 20 Flaws in Open-Source Software

Sources

Additional media

Similar Stories

Similar Stories