OpenAI Launches GPT-5.6-Cyber for Approved Cybersecurity Defenders
OpenAI introduces a gated cybersecurity model for exploit research and vulnerability analysis.
OpenAI has released GPT-5.6-Cyber, a new model built for cybersecurity work. The model is being distributed through a gated access tier called Daybreak Red, which is limited to approved security defenders.
The model is intended to support exploit development and vulnerability research. OpenAI positioned the release around a tightening defense window, arguing that threat actors will increasingly rely on AI to execute attacks at greater speed and scale, including in fully autonomous ways.
By restricting access to trusted defenders, OpenAI aims to reduce the potential for misuse while still giving security teams advanced tools. The company did not disclose what criteria determine eligibility for the Daybreak Red tier or when broader access might be considered.
The launch reflects a broader industry push to apply large language models to security operations, even as concerns persist about their dual-use nature. Further technical details about GPT-5.6-Cyber's capabilities were not included in the initial announcement.