Open-weight AI models can exploit the patch gap

Open-weight AI model GLM 5.3 can identify and exploit vulnerabilities in Firefox during the "patch gap"—the period between when security fixes are published and when they reach users—enabling attackers to compromise browsers for as little as $1,000 per version. The model successfully created malicious webpages that could steal files from users' computers without their interaction, exploiting security information available before patches were shipped. This demonstrates that widely available open-weight models now enable relatively unsophisticated attackers to exploit software vulnerabilities that previously required significant expertise or access to frontier AI systems.