Daybreak
OpenAI's gated programme that gives vetted security professionals access to models with the usual safety filters relaxed.
Daybreak is how OpenAI hands cybersecurity capabilities to people it has checked first. Normal models refuse requests that look like hacking, which is sensible, but it also blocks the defenders whose job is exactly that. Daybreak removes those blocks for approved individuals and organisations, behind identity verification, monitoring, legal attestations and, from September 2026, mandatory hardware security keys.
Since August 2026 it comes in two levels. Daybreak Blue gives the ordinary frontier model with defensive work unblocked, which OpenAI recommends for most defenders. Daybreak Red adds purpose-trained models such as GPT-5.6-Cyber for vulnerability research and exploit development. The gap is large: on OpenAI’s internal test of advanced requests, the specialist model answers 95 percent where the standard one answers 1.5 percent.