OpenAI has unveiled Daybreak, a cyber defense framework focused on building software that is secure by default through the integration of AI-powered security across the entire development lifecycle.

The system uses AI to help security teams reason across large codebases, uncover subtle flaws, test fixes, and shorten the time between identifying and resolving vulnerabilities. With Daybreak, OpenAI aims to make software continuously more secure by integrating defense into standard development workflows.

Daybreak combines OpenAI models with Codex-driven agentic tooling and external security partners to support capabilities such as secure code review, threat modeling, dependency analysis, and remediation guidance. These tools are intended to strengthen the entire software security lifecycle.

Given the dual-use nature of advanced AI, OpenAI says Daybreak includes safeguards around verification, proportional controls, and accountability, and will be deployed iteratively with input from industry and government partners.