OpenAI released a suite of new cybersecurity tools and the full version of GPT-5.5-Cyber on June 22, 2026, aimed at automating the discovery and patching of software vulnerabilities at scale. The company expanded its Daybreak initiative, introducing Codex Security plugin updates and the Daybreak Cyber Partner Program. This initiative aims to address the shifting cybersecurity bottleneck from vulnerability discovery to the speed of patching. Codex Security, which entered research preview in March, has scanned over 30 million commits across 30,000 codebases, with human reviewers marking 70,000 fixes and 500,000 findings resolved automatically.

The full version of GPT-5.5-Cyber demonstrates improved performance in security-focused benchmarks. On CyberGym, it achieved 85.6%, surpassing the 81.8% score of GPT-5.5. Results on ExploitGym reached 39.5% compared to 25.95%, and on SEC-bench Pro, it attained 69.8% against 63.1%. Additionally, OpenAI launched 'Patch the Planet,' a project founded with Trail of Bits and HackerOne to assist over 30 open-source projects, including Go, Python, and cURL, in automating security fixes. OpenAI is also providing Trusted Access for Cyber to international partners, including institutions in Japan, Germany, France, Canada, Australia, the Republic of Korea, and the EU.

OpenAI released a suite of new cybersecurity tools and the full version of GPT-5.5-Cyber on June 22, 2026, aimed at automating the discovery and patching of software vulnerabilities at scale. The company expanded its Daybreak initiative, introducing Codex Security plugin updates and the Daybreak Cyber Partner Program. This initiative aims to address the shifting cybersecurity bottleneck from vulnerability discovery to the speed of patching. Codex Security, which entered research preview in March, has scanned over 30 million commits across 30,000 codebases, with human reviewers marking 70,000 fixes and 500,000 findings resolved automatically.

The full version of GPT-5.5-Cyber demonstrates improved performance in security-focused benchmarks. On CyberGym, it achieved 85.6%, surpassing the 81.8% score of GPT-5.5. Results on ExploitGym reached 39.5% compared to 25.95%, and on SEC-bench Pro, it attained 69.8% against 63.1%. Additionally, OpenAI launched 'Patch the Planet,' a project founded with Trail of Bits and HackerOne to assist over 30 open-source projects, including Go, Python, and cURL, in automating security fixes. OpenAI is also providing Trusted Access for Cyber to international partners, including institutions in Japan, Germany, France, Canada, Australia, the Republic of Korea, and the EU.