Google’s DeepMind has introduced CodeMender, an AI-powered agent that autonomously identifies and patches software vulnerabilities. This innovation aims to keep pace with AI’s ability to uncover flaws, enhancing security across open source projects and beyond. #DeepMind #CodeMender
Keypoints
- Google’s DeepMind launched CodeMender, an AI agent for finding and fixing security vulnerabilities.
- CodeMender uses advanced analysis techniques like static/dynamic analysis and fuzzing to identify flaws.
- The system employs multi-agent teams to verify and ensure the safety of patches before deployment.
- Over six months, CodeMender has submitted 72 security patches for large open source projects.
- DeepMind emphasizes cautious review of patches, ensuring fixes do not cause regressions or new issues.
Read More: https://www.securityweek.com/google-deepminds-new-ai-agent-finds-and-fixes-vulnerabilities/