At a Glance · AI-generated

Open-weight AI models demonstrated significant cybersecurity vulnerabilities over two days in late October. Researchers found that the open-weight model GLM 5.3 independently constructed complete Firefox exploit chains without human assistance in under 48 hours, exploiting the window between vulnerability disclosure and patch deployment. The findings raised questions about the cyber risks posed by publicly available AI models and their potential to accelerate exploitation of unpatched security flaws.