OpenAI Daybreak

What it does

Daybreak is an OpenAI programme for approved security work with two access tiers. Daybreak Blue gives you the general frontier models such as GPT-5.6 Sol with relaxed safeguards, for example for vulnerability triage, secure code review, malware analysis, incident response and patch validation. Such requests are refused less often, though refusals still happen. Daybreak Red adds access to the specialised GPT-5.6-Cyber model, meant for explicitly authorised tasks such as reproducing vulnerabilities, proof-of-concept validation, penetration testing and red teaming. You can use Daybreak in ChatGPT, in Codex, for example together with the Codex Security agent, and via the API.

Who it suits

Security teams, pentesters and development teams that build security checks into their routine. As an individual, you need a paid plan, Advanced Account Security and a FIDO2 hardware security key. Companies apply through an enterprise request form.

Data protection for businesses

By default, OpenAI does not use data from ChatGPT Business, Enterprise, Edu or the API for training. Storage in Europe (EEA and Switzerland) is available to new Enterprise and Edu customers and to eligible API customers. Codex on the web is excluded. Daybreak approval does not automatically include Zero Data Retention. OpenAI signs a data processing agreement for Business, Enterprise and the API on request.

Limits

Approval is not guaranteed. Red needs separate approval. Use is limited to internal work, not customer projects. Daybreak does not isolate environments: you set up the sandbox and permissions yourself.