OpenAI has launched two access categories under its Daybreak cybersecurity initiative to provide certified defenders with access to AI models for authorized security work.
The move comes as threat actors are anticipated to employ AI to launch cyberattacks with increased speed and scale. There are two tiers: Daybreak Blue and Daybreak Red.
Daybreak Blue grants access to general-purpose frontier models, such as GPT-5.6 Sol, with precautions customized to authorized defensive security operations. It facilitates processes like as vulnerability identification, secure code review, malware analysis, incident response, and patch validation.
Daybreak Red offers cybersecurity-focused models for authorized vulnerability research, exploit validation, and security testing.
OpenAI has also released GPT-5.6-Cyber via Daybreak Red. The model, which is based on GPT-5.6 Sol, is taught to do specialized cybersecurity tasks like as detecting zero-day vulnerabilities and constructing attack chains. It also aims to decrease refusals for higher-risk, dual-use cybersecurity projects.
The company is collaborating with Daybreak partners and the open-source community to identify and resolve the vulnerabilities.
According to its Preparedness Framework, GPT-5.6-Cyber meets the High criterion for cybersecurity capacity but not the Critical requirement. The model outperformed GPT-5.6 Sol on several specialized cybersecurity tasks, but it did not attain the Critical level.
Daybreak Blue and Daybreak Red are only available to recognized persons and organizations carrying out authorized work. Identity verification, account security, monitoring, approved-use limits, and legal attestations all serve to manage access.
Beginning September 1, 2026, the business will require individual Daybreak accounts to utilize hardware security keys, and it is developing on additional security and monitoring methods. It also encourages Daybreak customers who use Codex to switch from full-access mode to auto-review mode, which checks for elevated permissions before executing them.
OpenAI advises organizations to perform security procedures in controlled settings, monitor agent activity, and precisely specify the systems and operations that agents are permitted to access.
