OpenAI release GPT-6 Cyber, first open to approved security teams

The news of November 6thOpenAI releaseCybersecurity model GPT-6 Cyber, the first to be opened through API and Codex to selected Daybreak Red clients. OpenAI rated its capabilities in cybersecurity at the highest "key" level, and previous GPT-5.6 Cyber rated it "high"。

In the Linux kernel study of OpenAI, GPT-6 Cyber reproduced 1724 loopholes and developed operational local rights options for 948 of them. The team disclosed 304 of these key findings to defenders in line with the Linux community guidelines and provided candidate patches。

The model also found 81 previously unknown high-risk loopholes in Firefox and assisted in the development of patches. OpenAI states that Mozilla has repaired many of these loopholes。