OpenAI has launched Daybreak, a security-focused AI initiative designed to find and fix software vulnerabilities before attackers get the chance. Built on top of the Codex Security AI agent that debuted in March, Daybreak maps an organization's codebase to build a threat model, traces likely attack paths, validates potential vulnerabilities, and automates detection of the highest-risk ones.
The timing is pointed. Anthropic announced Claude Mythos just over a month ago, a security-oriented AI model the company said was too dangerous to release publicly. Anthropic kept it tightly restricted within its own Project Glasswing initiative. OpenAI's move into the same space signals that AI-assisted offensive and defensive security tools are becoming a serious competitive battleground among the major labs.
Codex, the underlying agent powering Daybreak, was already positioned as a coding assistant before OpenAI sharpened its focus toward security applications. With Daybreak, the company is now pitching it directly to organizations looking to get ahead of software exploits rather than respond to them after the fact. The full scope of Daybreak's availability and pricing has not yet been disclosed.
