OpenAI rolls out more capable version of its cybersecurity model

On June 22, 2026, OpenAI announced an update to its GPT-5.5-Cyber model, making it 'more permissive and more capable for advanced, authorized cybersecurity work.' The company also launched the OpenAI Daybreak Cyber Partner Program, designed to provide partners with early access to the updated model.

The updated model builds on OpenAI's existing GPT-5.5-Cyber, which was initially released for cybersecurity applications. The 'more permissive' designation indicates that the model has fewer restrictions on its outputs, allowing for a wider range of security-related tasks. The Daybreak Cyber Partner Program is intended to facilitate collaboration with external organizations in developing cybersecurity solutions using the model.

The update comes amid heightened attention to AI cyber risks.