OpenAI introduced Daybreak, a cybersecurity push built around AI tools for finding and fixing software vulnerabilities.
The initiative includes Codex Security and GPT-5.5-Cyber, positioning frontier models as assistants for validation, patching, and security operations rather than only code generation.
The announcement lands in a sensitive area for AI labs: the same model capabilities that can help defenders can also raise concerns about offensive cyber use, so evaluation and deployment controls will matter.