Google has announced that its AI-powered bug hunter, “Big Sleep,” has discovered 20 security vulnerabilities in widely used open source software such as FFmpeg and ImageMagick. This marks the first public results from Big Sleep, which was created by Google’s DeepMind and Project Zero teams. While details about the vulnerabilities are being withheld until they are fixed, Google’s announcement is notable as it demonstrates that AI tools can independently find and reproduce real software flaws, with a human expert only involved for verification and reporting. This achievement highlights the growing effectiveness of large language model (LLM)-based tools for automated vulnerability discovery. However, the technology is not without challenges, as some in the industry have experienced false positives or “hallucinated” bug reports from AI tools. Google’s approach—using humans to validate findings—aims to address this concern. Overall, the development points to significant potential for AI to improve cybersecurity, while underscoring the need for oversight to ensure accuracy.
Read the News behind this article:Link
