Aug 132026 More than half of AIgenerated patches are broken New research shows AI models like ChatGPT and Claude frequently fail to fix cybersecurity bugs, often introducing fresh code vulnerabilities instead.
