OpenAI has entered the cybersecurity arms race with Daybreak, a full-stack platform launched May 11, 2026, designed to automate digital defense for security teams.
The platform operates in four stages: building threat models by scanning codebases, prioritizing vulnerabilities by real-world exploitability, validating threats in controlled sandbox testing, and generating contextual patches for human review.
At its heart is Codex Security, OpenAI’s specialized code-analysis model, which during beta resolved over 3,000 high-severity vulnerabilities.
Daybreak offers three tiers. The base layer runs on GPT-5.5 for general use. The second, GPT-5.5 Trusted Access for Cyber, integrates with partners like Zscaler for defensive workflows. The top tier, GPT-5.5-Cyber, is built for offensive red-teaming and requires phishing-resistant authentication by June 1.
Anthropic has also launched competing offerings-Project Glasswing and Claude Mythos 2-both targeting similar enterprise security use cases.