Skip to main content

Quick Overview

This video is an administrative configuration tutorial presented by Dan Shanahan, Cyber Deployment Engineer at OpenAI. It demonstrates the technical procedures for requesting, provisioning, and validating Daybreak Blue access for enterprise workspaces. The guide was produced to help administrators securely implement frontier model capabilities for cyber defense.

Key Points

  • 1.Daybreak Blue provides cybersecurity defenders with access to OpenAI frontier models under specialized safeguards.
  • 2.Access requires an organization to complete third-party business and identity verification using Persona before submitting an intended use form and business terms.
  • 3.Administrators can enable Daybreak Blue globally across a workspace or restrict it to designated users through role-based access control overrides.
  • 4.Users can verify their access within Codex by checking the Daybreak Access status in settings and switching the Daybreak toggle in the model selector.
  • 5.API integration requires creating a separate project in the OpenAI developer platform, enabling Daybreak model access in project settings, assigning project members, and generating dedicated secret keys.

Summary

Dan Shanahan, Cyber Deployment Engineer at OpenAI, presents a technical walkthrough demonstrating how to request, configure, and validate Daybreak Blue access across an organization. Daybreak Blue provides frontier model capabilities tailored for cybersecurity defense. The setup process begins inside the OpenAI Admin Console under the Trusted Access navigation tab, which houses business verification and application workflows.

To initiate access, administrators select the business verification option, redirecting them to a third-party verification portal hosted by Persona. Depending on geographic location, the user must present a valid government-issued identification document to verify their identity and submit company registration information and business details. After the Persona flow is completed, the administrator returns to the console to fill out an intended use form, accept the Daybreak business terms, and submit the application. The request enters an in-review status before transitioning to verified and active.

After approval, administrators manage availability by navigating to the Models section of the Admin Console. While Daybreak Blue can be enabled as a workspace default for every user, organizations typically require granular role-based access control. Shanahan demonstrates opening the Roles section, selecting a custom Daybreak role, creating an edit override, and toggling Daybreak Blue from the default off state to on. Saving these changes restricts model access strictly to users assigned to that role.

To verify that permissions have propagated to end users, Shanahan opens the Codex interface and navigates to the user settings menu. Under the General settings pane, a banner confirming Daybreak Access shows that expanded cyber capabilities are active for approved security work. Returning to the main chat interface, the model selector exposes a dedicated Daybreak toggle alongside models such as GPT-6 Sol, allowing the operator to engage frontier models with Daybreak controls enabled.

For automation pipelines and external test harnesses, access must also be established at the API level via platform.openai.com. In the developer platform settings, Shanahan notes that the default project cannot have Daybreak Blue enabled directly. Administrators must create a new project, enter project settings, and toggle on Daybreak model access to permit eligible models, including gpt-5.3-codex and gpt-daybreak-blue-latest. Once the project settings are saved, the administrator adds authorized users to the project member list and issues a new secret API key scoped to that project.

Requesting Daybreak Blue Verification

Administrators navigate to the Trusted Access section of the OpenAI Admin Console to begin the application. The organization must complete an identity and business verification flow handled through Persona, requiring government identification and corporate entity details. After verification is submitted, applicants complete an intended use form, review and accept the Daybreak business terms of service, and submit the request for review.

Configuring Workspace and Role-Based Model Access

Once an organization's request is approved and marked active, administrators navigate to the Models tab within the Admin Console. Daybreak Blue can be enabled as a workspace default for all members or scoped using role-based access control. Creating an override within a designated role enables administrators to toggle Daybreak Blue access specifically for users assigned to that role without exposing it workspace-wide.

Validating Access Inside Codex

Users assigned Daybreak permissions can confirm their status by opening Codex settings and checking the General tab. An active Daybreak Access banner indicates valid authorization for expanded cyber capabilities. Users can then toggle the Daybreak control directly inside the Codex model selector to interact with compatible frontier models.

Provisioning API Keys for Automation

To build automations or run evaluation harnesses, administrators access the OpenAI developer platform. Because Daybreak Blue cannot be enabled on the default project, administrators must create a separate project, enable the Daybreak model access toggle under project settings, add team members to the project, and generate project-scoped API keys.

The Bottom Line

The video establishes the end-to-end administrative sequence required to request Daybreak Blue, enforce role-based access controls, validate end-user availability in Codex, and generate project-specific API keys. It lands on a complete setup for both interactive and programmatic cyber defense workloads. It leaves operational model performance metrics and specific approval timelines unaddressed.

FAQ

What is Daybreak Blue and what is it used for?

Daybreak Blue is an OpenAI offering that brings frontier model access and expanded cyber capabilities to cybersecurity defenders for approved security work.

How do administrators submit an application to request Daybreak Blue access?

Administrators navigate to the Trusted Access section in the OpenAI Admin Console, complete identity and business verification via Persona, fill out an intended use form, review and accept the Daybreak business terms, and submit the application for review.

How can an administrator restrict Daybreak Blue access using role-based access control?

In the Admin Console, administrators navigate to the Roles tab under Models, select the target role, open the edit override interface, switch Daybreak Blue to on, and save the changes so only members of that role receive access.

Where does an individual user verify their Daybreak Blue authorization status in Codex?

A user opens Codex, navigates to user settings under the General tab, and checks for the Daybreak Access banner confirming that expanded cyber capabilities are enabled.

Why must administrators create a new project on platform.openai.com for Daybreak Blue API keys?

Daybreak Blue model access cannot be enabled on the default project in the OpenAI developer platform, requiring administrators to create a dedicated project before toggling on Daybreak model access and issuing secret keys.

Worth watching for

Security operations teams, systems administrators, and cyber engineers responsible for deploying and managing OpenAI frontier models for cybersecurity workloads.

  • openai
  • daybreak-blue
  • cybersecurity
  • codex
  • access-control
  • api-keys