OpenAI's Daybreak Brings GPT-5.5 and Codex to Cyber Defenders - With a New Model for Red Teams

OpenAI has unveiled Daybreak, its first dedicated cybersecurity programme that integrates GPT-5.5 and the Codex coding agent into a comprehensive defence platform. The initiative marks OpenAI's entry into enterprise security, positioning AI models as both development tools and security guardians within the same workflow. Daybreak enables security teams to conduct code reviews, threat modelling, dependency analysis, and patch validation directly within the development cycle, rather than treating security as an afterthought. This shift-left approach promises to catch vulnerabilities earlier in the development process, when fixes are typically less costly and disruptive. Access to Daybreak operates through three distinct tiers. Standard GPT-5.5 serves general security tasks, whilst a Trusted Access tier caters to verified security professionals. The most powerful offering, GPT-5.5-Cyber, provides enhanced capabilities for authorised red team operations and penetration testing, though access remains strictly limited. OpenAI has acknowledged the inherent dual-use challenge: the same AI capabilities that identify vulnerabilities could theoretically be used to exploit them. The tiered access structure represents an attempt to ensure the most powerful tools remain in defenders' hands. This launch also intensifies competition with Anthropic's Mythos platform, which has been targeting similar enterprise security use cases, establishing cybersecurity as a new battleground in the frontier AI race.