OpenAI announced GPT-6 Cyber for a select group of approved Daybreak Red customers.
Restricted access and authorised scope remain central to evaluating security-specialised models.
What happened
OpenAI’s 5 November announcement introduces GPT-6 Cyber for authorised security research, initially through the API and Codex for a select group of approved Daybreak Red customers. OpenAI reports stronger cybersecurity and scope-following performance; these are developer claims, not an independent assessment. Its system card describes identity verification and organisational security requirements, including single sign-on, multi-factor authentication and API IP allowlisting.
Why it matters: Restricted access and authorised scope remain central to evaluating security-specialised models.
Sources: Primary source, “GPT-6 Cyber announcement”; Primary source, “GPT-6 Cyber System Card - OpenAI Deployment Safety Hub”