Nov 5, 2026
OpenAI releases a cybersecurity-specialized variant of GPT-6 Astra for approved security researchers and organizations.
Access requires Daybreak Red approval. OpenAI classifies the model as Critical in cybersecurity and reports improvements in vulnerability discovery, exploit validation and respecting authorized task boundaries. The model release follows the separate Linux audit disclosure.