OpenAI Expands Daybreak Program with Codex Security Plugin and GPT-5.5-Cyber Model.

OpenAI announced a major expansion of its Daybreak initiative aimed at helping security teams identify and patch software vulnerabilities at machine speed. The update includes a Codex Security plugin, an improved GPT-5.5-Cyber model, and a new partner program bringing advanced defensive AI capabilities to more than 20 security companies.

The Codex Security plugin integrates directly into Codex, following recent updates that brought agent capabilities to the platform's development environment. Teams can now run deep scans, validate findings, trace attack paths, build threat models, and generate codebase-specific patches for human review without leaving their workflow. The plugin also exports findings into existing security tools, making it easier to incorporate AI-assisted vulnerability management into established processes.

GPT-5.5-Cyber has been updated to be more permissive and more capable for advanced, authorized cybersecurity work. According to OpenAI's announcement, it delivers stronger performance on defensive workflows including tracing vulnerable code, validating issues, developing patches, and preparing evidence for human review.

The updated model represents what OpenAI describes as its most capable cyber model yet. Unlike general-purpose models, GPT-5.5-Cyber is specifically optimized for security professionals conducting authorized defensive operations. The increased permissiveness suggests the model can now assist with more advanced security research tasks while maintaining guardrails against misuse.

The OpenAI Daybreak Cyber Partner Program allows participating security software and services providers to use GPT-5.5 with Trusted Access for Cyber in the products they deliver to customers. This approach enables end users to benefit from advanced defensive AI capabilities while keeping direct model access limited to trusted partners.

More than 20 leading security companies have joined the program at launch. By embedding these capabilities into existing security products, OpenAI aims to democratize access to AI-powered vulnerability detection and patching without requiring every organization to integrate directly with the models themselves.

The program also includes Patch the Planet, an initiative working with open-source maintainers to secure critical projects. Given that open-source software underpins much of the internet's infrastructure, addressing vulnerabilities in widely-used libraries and frameworks could have outsized impact on overall software security.

OpenAI positioned the expansion as a shift in focus from simply identifying more vulnerabilities to actually making software safer. The company stated its goal is to move toward a world of safer software and cyber resilience, rather than just better bug detection.

Security teams have long grappled with alert fatigue and the challenge of prioritizing which vulnerabilities to address first. By automating not just detection but also validation, patch generation, and evidence preparation, the Codex Security plugin could help teams move faster through the entire remediation lifecycle. However, OpenAI emphasized that patches are generated for human review, maintaining human oversight in the security workflow.