OpenAI announced the GPT‑5.6‑Cyber model, available through Daybreak Red access. It is designed to handle advanced cybersecurity tasks with fewer refusals. The company states that it has conducted an internal test and outperformed its general-purpose models. The company offered GPT-5.6-Cyber access to trusted customers, who used the model to improve their defensive cybersecurity workflows. The early-access partners included SpecterOps, SentinelOne and Palo Alto Networks.
