OpenAI announced expansion of its Daybreak cybersecurity initiative and introduced GPT-5.6-Cyber. The model targets advanced authorized tasks including vulnerability research, exploit validation, and security testing. Access is restricted to approved defenders, with Daybreak Red for higher-risk work and Daybreak Blue for broader defensive activities such as malware analysis and incident response. Company posts state the goal is placing frontier capabilities with trusted users before attackers gain similar tools. One reply questioned why the model was not deployed to protect OpenAI's own network after reported attacks on it and Hugging Face.