OpenAI has unveiled Daybreak, a new cybersecurity platform that combines its frontier models with Codex as an agentic framework. The tool lets security teams generate patches, test them, and verify fixes directly inside their codebase.
To balance power with abuse risk, Daybreak splits model access into three strict tiers:
- Standard GPT-5.5 with normal safety guardrails.
- GPT-5.5 with "Trusted Access" — more refined guardrails for everyday defense tasks like malware analysis and vulnerability triage.
- GPT-5.5-Cyber — the most open version, given only to authorized users for red-blue teaming (simulated attacks) and penetration testing.
By embedding Codex as an agent framework into development workflows, Daybreak aims to push large language models beyond vulnerability discovery and into automated remediation. The platform is already open for enterprise vulnerability scanning requests, and Cloudflare is the first publicly named test customer.