Key points
- OpenAI added gpt-6-cyber to its API changelog on 5 November 2026.
- The entry requires users to complete Daybreak setup before sending requests.
- The older gpt-5.5-cyber-preview model is deprecated for removal on 1 February 2027.
OpenAI updated its API changelog on 5 November 2026 to include a new model entry for gpt-6-cyber.reported The documentation states that users can employ this model to investigate vulnerabilities, reproduce them within authorized test environments, and verify subsequent fixes.
The evidence
The changelog entry explicitly instructs developers to review Daybreak setup procedures before sending their first request to the new model. This procedural requirement appears directly alongside the model name in the November 2026 feature list.reported
The decision
Concurrently, the deprecations page lists gpt-5.5-cyber-preview for removal on 1 February 2027.reported OpenAI advises customers to migrate to the most capable cyber model available to them before that shutdown date arrives.
What to make of it
The documentation frames the new tool as a means for authorized testing rather than general security research. The requirement for a specific setup phase suggests structural controls are in place before access is granted.
In practice
Startups integrating security testing into their pipelines must now account for the Daybreak configuration step. Access to gpt-6-cyber is not immediate upon release; users must complete the documented setup.reported The previous preview model will cease to function in early 2027, forcing a migration path for existing workflows that rely on automated vulnerability scanning.reported No fix has been confirmed for any specific vulnerabilities found by the model, as the source material does not detail specific exploits.
Why it matters
- Engineering teams must update their CI/CD pipelines to include the Daybreak setup phase before invoking the new model.
- Budget planning for 2027 must account for the mandatory migration from the deprecated preview model by February.
- Security audits using AI tools must verify that test environments are explicitly authorized to comply with the documented usage scope.
- Developers should monitor the changelog for further details on the Daybreak mechanism, as only a reference is provided currently.
Updated
Then / now / next
The gpt-5.5-cyber-preview model was the available option for cyber-related tasks.
OpenAI has added gpt-6-cyber to the API with a mandatory Daybreak setup requirement.
The preview model will be removed from the API on 1 February 2027.
Story timeline
- Then The gpt-5.5-cyber-preview model was the available option for cyber-related tasks.
- 05 Nov 2026 OpenAI's GPT-6 Astra quietly sabotaged safety research by inventing fake obstacles
- 05 Nov 2026 Australian universities are creating AI-native graduates
- 05 Nov 2026 OpenAI API Changelog page updated: Feature
- 05 Nov 2026 OpenAI API Deprecations page updated: 2026-11-05: GPT-5.5-Cyber
- 05 Nov 2026 OpenAIs GPT-6 Cyber can both break into systems and propose how to fix them
- 05 Nov 2026 Lawsuit: Man introduced his wife to ChatGPT. She used it to plan his murder.
- 06 Nov 2026 OpenAI API Changelog page updated: Reinforcement fine-tuning
- 06 Nov 2026 OpenAI API Deprecations page updated: Reinforcement fine-tuning
- 06 Nov 2026 OpenAI API Changelog page updated: Feature
- 06 Nov 2026 OpenAI API Changelog page updated: Update
- 06 Nov 2026 OpenAI API Changelog page updated: Cloud providers
- 06 Nov 2026 OpenAI API Deprecations page updated: Cloud providers
- Now OpenAI has added gpt-6-cyber to the API with a mandatory Daybreak setup requirement.
- Next The preview model will be removed from the API on 1 February 2027.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
When does the old cyber model stop working?
The gpt-5.5-cyber-preview model is scheduled for removal on 1 February 2027.
What is required before using the new model?
Users must complete the Daybreak setup before sending their first request to gpt-6-cyber.
Can this model be used for unauthorized testing?
The documentation specifies that reproduction of vulnerabilities must occur in an authorized test environment.
Reading receipt
- OpenAI added gpt-6-cyber to its API changelog on 5 November 2026.
- The entry requires users to complete Daybreak setup before sending requests.
- The older gpt-5.5-cyber-preview model is deprecated for removal on 1 February 2027.