AI can turn security fixes into working exploits before users receive an update
Summary
Epoch AI reports that GLM 5.3 open-weight models can exploit the patch gap to compromise six Firefox releases, at roughly $1,000 per version. The findings highlight growing risks as open AI models gain offensive capabilities, emphasizing the need for faster, private security fixes and stronger defense measures. The study also notes patch deployments and browser/vendor responses, underscoring implications for software supply chain security.