OpenAI AI Pipeline Generates 948 Linux Root Escalation Exploits in 30-Hour Run

OpenAI has detailed a new AI pipeline that utilized an upcoming cyber-focused AI model to identify, reproduce, and generate 948 local privilege escalation (LPE) exploits in Linux within a 30-hour run. This process, which cost approximately $1.2 million in API credits, involved identifying high-value audit targets, dynamically validating vulnerabilities, assessing exploitability, and generating patches. The findings build upon previous discoveries by OpenAI's GPT-5.5-Cyber model.