Who can use GPT-5.5-Cyber?

OpenAI is keeping GPT-5.5-Cyber on a limited release for verified defenders only, with extra monitoring and controls. It is more permissive than the standard GPT-5.5 for authorized security work, but it is not commercially available — access requires approval. A standard GPT-5.5 with Trusted Access for Cyber is available to general users for less sensitive defensive work.

What does Codex Security actually do?

Codex Security scans a repository, checks whether a flagged code path is reachable, writes a patch, and tests it before a human reviewer signs off. OpenAI says the plugin has covered more than 30 million commits across 30,000+ codebases since its March 2026 preview, with over 500,000 findings closed automatically and another 70,000 marked resolved by humans.

What is Patch the Planet?

Patch the Planet is a Daybreak initiative founded with Trail of Bits that funds researchers to help open-source maintainers fix bugs. More than 30 widely-used projects are signed up at launch, including cURL, Go, and Python. Maintainers receive ChatGPT Pro, conditional Codex access, and API credits to help work through their backlog.

How does GPT-5.5-Cyber compare to GPT-5.5?

GPT-5.5-Cyber scores 85.6% on CyberGym, a vulnerability-reproduction benchmark, against 81.8% for the standard GPT-5.5. It also jumps from roughly 26% to nearly 40% on ExploitGym. OpenAI says the model is both more capable and more permissive than its general models for authorized security work, which is why access is gated.

OpenAI · 2026-06-22 · major

OpenAI Daybreak — GPT-5.5-Cyber and Patch the Planet go live

OpenAI expands its Daybreak security program with the full release of GPT-5.5-Cyber, an updated Codex Security plugin, a partner program with CrowdStrike, Sophos, and Fortinet, and Patch the Planet, an open-source fix-funding effort with Trail of Bits.

OpenAI Daybreak cybersecurity program expansion announcement — ITPro / OpenAI

GPT-5.5-Cyber, Codex Security, and Patch the Planet — OpenAI bets the bottleneck has moved from finding flaws to shipping fixes.

Key specs

Cyber gym score	85.6%
Commits scanned	30M+
Findings closed	500K+

Quick facts

Maker	OpenAI
What's new	GPT-5.5-Cyber full release, Codex Security update, partner program, Patch the Planet
GPT-5.5-Cyber access	Limited release for verified defenders
Codex Security coverage	30M+ commits across 30K+ codebases since March preview
Patch the Planet seed projects	30+, including cURL, Go, Python
Partner program members	CrowdStrike, Sophos, Fortinet, Palo Alto, Cisco, Cloudflare
Announced	June 22, 2026

What is it?

Daybreak adds three new pieces to OpenAI's security stack: GPT-5.5-Cyber moves from preview to a limited full release for verified defenders, the Codex Security plugin gets an upgrade aimed at validating and patching flaws in production code, and Patch the Planet funds open-source maintainers to actually ship those fixes. A new partner program lets CrowdStrike, Sophos, Fortinet, Palo Alto, Cisco, and Cloudflare embed Daybreak models in their products.

How does it work?

The Codex Security plugin scans a repository, checks whether a flagged vulnerability is reachable from real entry points, writes a patch, and runs tests before a human reviewer signs off. GPT-5.5-Cyber is tuned for that loop — OpenAI reports 85.6% on CyberGym and a jump to nearly 40% on ExploitGym, compared to 81.8% and ~26% for the general GPT-5.5. Patch the Planet pairs the model with researchers and gives maintainers ChatGPT Pro plus API credits to work through bug backlogs.

Why does it matter?

Five Eyes intelligence agencies warned on the same day that frontier AI will reshape offensive cyber capabilities in months, not years. OpenAI is arguing the right defense is to make patching as cheap as finding bugs — the long-standing bottleneck for open-source projects that everyone depends on. If Codex Security's numbers hold up, individual maintainers of cURL, Go, and Python suddenly get a security engineer who works full time for them.

Who is it for?

security teams, OSS maintainers, platform vendors

Frequently asked questions

Who can use GPT-5.5-Cyber?: OpenAI is keeping GPT-5.5-Cyber on a limited release for verified defenders only, with extra monitoring and controls. It is more permissive than the standard GPT-5.5 for authorized security work, but it is not commercially available — access requires approval. A standard GPT-5.5 with Trusted Access for Cyber is available to general users for less sensitive defensive work.
What does Codex Security actually do?: Codex Security scans a repository, checks whether a flagged code path is reachable, writes a patch, and tests it before a human reviewer signs off. OpenAI says the plugin has covered more than 30 million commits across 30,000+ codebases since its March 2026 preview, with over 500,000 findings closed automatically and another 70,000 marked resolved by humans.
What is Patch the Planet?: Patch the Planet is a Daybreak initiative founded with Trail of Bits that funds researchers to help open-source maintainers fix bugs. More than 30 widely-used projects are signed up at launch, including cURL, Go, and Python. Maintainers receive ChatGPT Pro, conditional Codex access, and API credits to help work through their backlog.
How does GPT-5.5-Cyber compare to GPT-5.5?: GPT-5.5-Cyber scores 85.6% on CyberGym, a vulnerability-reproduction benchmark, against 81.8% for the standard GPT-5.5. It also jumps from roughly 26% to nearly 40% on ExploitGym. OpenAI says the model is both more capable and more permissive than its general models for authorized security work, which is why access is gated.

Try it

openai.com/daybreak