
OpenAI releases GPT-6 Cyber model, says it is less likely to hack companies
The story in brief
OpenAI released GPT-6 Cyber, its most advanced cybersecurity model, which is less likely to hack companies. Researchers used it to find 81 previously unknown Firefox vulnerabilities, with another 30 issues flagged as more severe. The model completed 95% of tasks on RemoteExploitBench, with Fouad Matin noting its improved performance.