OpenAI structures its cybersecurity offering around Daybreak and Codex Security.

OpenAI launches Daybreak and Codex Security, integrating GPT-5.5 models into defensive workflows with partners like Cloudflare, Cisco, and CrowdStrike.

OpenAI is formalizing its strategy under the name Daybreak to integrate its models into defensive cybersecurity workflows. The central idea: to combine the intelligence of the models with an agentic harness (Codex) and an ecosystem of security partners so that resilience becomes a native property of the code, and not a patch applied after the fact.

The functional scope covers secure code review, threat modeling, patch validation, dependency risk analysis, detection, and remediation guidance, integrated into the daily development cycle. Codex Security, an operational component of the whole, builds an editable threat model from the repository, directs analysis towards realistic attack paths and high-exposure code, generates and tests patches directly within repositories under controlled access, then returns audit evidence to existing systems.

Three access levels structure the offering: GPT-5.5 by default with standard guardrails, GPT-5.5 with Trusted Access for Cyber for verified defensive work (vulnerability triage, malware analysis, detection), and GPT-5.5-Cyber with more permissive behavior for authorized specialized workflows such as red teaming and pentesting, accompanied by enhanced account and verification controls.

Several security partners are featured on the product page, including Cloudflare, Cisco, CrowdStrike, Palo Alto Networks, Oracle, Zscaler, Akamai, and Fortinet. OpenAI mentions an iterative deployment in the coming weeks with its industrial and governmental partners, accompanied by more advanced cybersecurity models.