OpenAI has introduced "GPT 5.6 Cyber," a new suite of models specifically engineered for cybersecurity applications such as vulnerability research, penetration testing, and incident response. The company has confirmed that these advanced capabilities will not be made available to general users due to potential security risks, instead restricting access to a select group of approved partners.
The new models are being integrated into existing security products, managed services, and customer engagements offered by these partners. OpenAI stated that this approach allows more defenders to identify critical vulnerabilities, validate their exploitability, and expedite remediation efforts by leveraging their frontier cyber models within established services.
Access to GPT 5.6 Cyber is provided through what OpenAI terms "Daybreak Access," which includes two distinct versions: Daybreak Blue and Daybreak Red. Daybreak Blue is designed for a broad spectrum of defensive security tasks, while Daybreak Red is intended for more specialized and closely governed operations.
OpenAI detailed several ways these models can assist security teams, including identifying vulnerabilities, assessing exploitability, pinpointing affected systems, developing fixes, and facilitating their deployment into production environments. Partners are expected to utilize these models for tasks such as vulnerability discovery and validation, red teaming, penetration testing, incident response, and remediation across various enterprise settings.
The list of initial partners includes major consulting firms like Accenture, IBM, Capgemini, Cognizant, EY, KPMG, PwC, NCC Group, and SpecterOps. Additionally, leading security vendors such as Palo Alto Networks, CrowdStrike, Cisco, Sophos, Akamai, Fortinet, and Cloudflare are also rolling out access to these new capabilities.
OpenAI emphasized that the underlying models remain under the control of the approved partners and are not directly transferred to customers. Safeguards implemented by partners may include identity verification, clearly defined testing scopes, logging, monitoring, and human oversight. Partners are responsible for collaborating with organizations to establish engagement boundaries, review findings, and apply their expertise before any actions are taken.
This controlled distribution strategy aims to provide enterprises with access to advanced AI-driven security capabilities without requiring them to develop their own specialized cyber AI infrastructure. Cybersecurity providers and consultancies can apply to join the Daybreak Cyber Partner program, while organizations interested in utilizing the technology can do so through participating security providers.
