Doyensec successfully exploited OpenAI's Codex model in the Coding Agent category at the PWN2OWN competition. However, the exploit was based on a previously disclosed vulnerability, leading the team to label it a "collision." They are now returning to further research. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Highlights potential vulnerabilities in AI coding assistants, underscoring the need for ongoing security research and patching.
RANK_REASON The cluster describes a successful exploit of an AI model at a security competition, which is a form of research milestone. [lever_c_demoted from research: ic=1 ai=1.0]
