Models

OpenAI Expands Daybreak: Full GPT-5.5-Cyber and Patch the Planet for Open Source

On June 22, 2026, OpenAI released updates under its Daybreak cybersecurity program: the full version of GPT-5.5-Cyber for trusted defenders (85.6% CyberGym vs 81.8% for GPT-5.5) and Patch the Planet, an initiative with partners including Trail of Bits to help open-source maintainers find, validate, and fix vulnerabilities with AI. Early results cited hundreds of issues and dozens of merged patches across major projects.

Tech Insights Reporter Jun 22, 2026 5 min read San Francisco, CA

TLDR

OpenAI on June 22, 2026 pushed its Daybreak cyber-defense stack forward with two primary posts: the full GPT‑5.5‑Cyber release for trusted defenders, and Patch the Planet, a program to accelerate vulnerability discovery and patching in open source. GPT‑5.5‑Cyber is reported at 85.6% on CyberGym versus 81.8% for GPT‑5.5. Trail of Bits engineers working full-time with Codex and the cyber model across 19 open-source projects had already identified hundreds of security issues and merged dozens of patches.

GPT-5.5-Cyber full release

From OpenAI’s Daybreak update:

Audience: Continued limited release to trusted defenders (not a free public offensive toolkit).
Positioning: More permissive for authorized cybersecurity work and more capable at finding and helping patch vulnerabilities, while retaining GPT‑5.5 general intelligence for long tasks.
Benchmark claim: 85.6% CyberGym vs 81.8% for standard GPT‑5.5 (vendor eval).
Lineage: Follows an earlier permissive-only preview; part of OpenAI’s broader Trusted Access for Cyber posture.

Patch the Planet

From OpenAI’s Patch the Planet post (Daybreak initiative):

Goal: Support open-source maintainers in finding, validating, and fixing vulnerabilities with AI assistance.
Partner example: Trail of Bits dedicated security engineers full-time with Codex and GPT‑5.5‑Cyber across 19 projects—hundreds of issues identified, dozens of patches merged, more under coordinated disclosure.
Early field notes (OpenAI-cited): fuzzing-lab construction in under a day that would “ordinarily take at least several weeks”; Linux kernel work including automated PoCs for pointer leaks and local privilege escalations among a larger issue set.

Together with related Codex security tooling, the package is OpenAI’s public answer to industry pressure—heightened after Anthropic’s Project Glasswing and the Fable/Mythos export fight—to put frontier cyber capability in defenders’ hands under access controls.

Why this story matters

Cyber capability is becoming the politically safest place to ship frontier uplift: governments care about critical-infrastructure defense, and labs can argue dual-use tools belong with verified responders. Daybreak’s full cyber model plus open-source patching scale is OpenAI’s bid to own that narrative—and to operationalize AI that writes exploits and patches—before rivals set the default stack.

Sources

OpenAI: “Daybreak: Tools for securing every organization in the world” / GPT‑5.5‑Cyber full update (openai.com/index/daybreak-securing-the-world/, June 22, 2026).
OpenAI: “Patch the Planet: a Daybreak initiative to support open source maintainers” (openai.com/index/patch-the-planet/, June 22, 2026).

Featured Image Alt Text

Shield icon with GPT-5.5-Cyber label over open-source code repositories and patch commit checkmarks.