GPT-5.6-Cyber refuses security researchers’ requests far less often
OpenAI's new GPT-5.6-Cyber model is less likely to refuse requests from security researchers, particularly for high-risk tasks, which enhances its utility for identifying vulnerabilities and building
