OpenAI Launches Daybreak AI Cybersecurity Platform
OpenAI on Monday launched Daybreak, a cybersecurity platform that combines GPT-5.5 models with the Codex Security agent framework to help organizations identify, validate, and resolve software vulnerabilities at scale. As a result, the company expanded its presence in the enterprise cybersecurity market and intensified competition with Anthropic’s Project Glasswing initiative.
Daybreak integrates GPT-5.5 with the Codex Security agent, which OpenAI first introduced in research preview in March under the codename Aardvark. The platform builds editable threat models directly from an organization’s code repositories. It then maps trust boundaries, identifies exposure points, and prioritizes vulnerabilities based on operational risk rather than generic scoring systems.
In addition, the platform validates vulnerabilities by executing proof-of-concept exploits inside sandboxed environments. It also generates contextual code patches that reduce the likelihood of software regressions. OpenAI reported that Codex Security helped resolve more than 3,000 critical and high-severity vulnerabilities during its beta phase.
Multiple Security Tiers and Industry Partnerships
OpenAI introduced three access tiers for Daybreak. The base tier supports general cybersecurity workflows with standard GPT-5.5 models. Meanwhile, the Trusted Access for Cyber tier provides expanded defensive capabilities with adjusted safeguards. The GPT-5.5-Cyber tier focuses on advanced security operations, including penetration testing, red teaming, and binary reverse engineering.
Furthermore, OpenAI partnered with several major cybersecurity companies to strengthen the platform’s adoption. Participating firms include Cloudflare, Cisco, CrowdStrike, Palo Alto Networks, Oracle, and Zscaler. Earlier this year, Zscaler integrated GPT-5.4-Cyber into its zero-trust security platform through OpenAI’s Trusted Access for Cyber program.
AI Security Competition Intensifies
The launch highlights the growing competition among AI companies in cybersecurity. Earlier this year, Anthropic introduced Project Glasswing and its Claude Mythos Preview model for defensive cyber operations. That initiative also attracted major technology partners and focused on AI-assisted security workflows.
Meanwhile, industry evaluations showed GPT-5.5 and Anthropic’s Mythos Preview performing at similar levels across advanced cyber tasks. At the same time, cybersecurity firms warned that frontier AI systems continue to accelerate attack timelines, increasing the need for automated AI-driven defense platforms.
OpenAI also confirmed that all approved Daybreak users and organizations must implement phishing-resistant authentication by June 1, 2026.
