According to CoinMeta, as disclosed on the official website of OpenAI, the company has established an autonomous security research pipeline that covers vulnerability discovery, dynamic verification, exploitability assessment, and patch generation. During a test run centered on a newly upcoming network model, the pipeline scanned and verified targets related to the Linux kernel, generating 948 vulnerabilities that could be exploited to achieve local privilege escalation (LPE) on common Linux releases. The cost of these exploits amounted to approximately $1.213 million, which is about $1,279 per exploit. In total, the pipeline identified and analyzed 2,308 potential vulnerabilities, dynamically reproduced 1,724 of them, and also uncovered 291 cases of information leakage. OpenAI indicates that critical vulnerability patches have been prioritized for reporting in accordance with the recommendations of the Linux community, and efforts are continuing to advance the reporting of the remaining vulnerabilities in collaboration with Patch, the, and Planet partners.