OpenAI unveils GPT-6 Cyber, a cybersecurity-focused AI model it says is designed to help authorized researchers find software vulnerabilities, develop exploits, and support fixes. Multiple outlets report that OpenAI says the model helped identify 81 high-severity vulnerabilities in Firefox and worked on exploits for 948 distinct Linux kernel flaws.
Several reports emphasize improvements over earlier versions. The Hacker News and SecurityWeek say OpenAI frames GPT-6 Cyber as more capable at finding and fixing vulnerabilities while better staying within boundaries set for authorized testing. Help Net Security and TechGenyz add that OpenAI tested the model in controlled security settings and that it reportedly uncovered issues earlier systems missed. Digital Trends and NDTV focus on timing and defensive framing—discovering flaws before attackers can exploit them—while Mashable highlights OpenAI’s claim that the model has already uncovered thousands of exploits.
Across coverage, outlets largely agree on the headline figures (81 Firefox flaws and 948 Linux exploit-related results), but differ in what they foreground, such as benchmarks (NDTV), the relationship to prior AI models (SecurityWeek, The Hacker News), and the scope implied by “thousands of exploits” (Mashable).