OpenAI has introduced GPT-5.5-Cyber, an advanced artificial intelligence model engineered to enhance vulnerability detection and automate patching processes at an unprecedented speed. This launch is a pivotal part of OpenAI’s Daybreak initiative, which strives to democratize cybersecurity resources for verified organizations globally.
Unmatched Cybersecurity Performance
GPT-5.5-Cyber excels in three key cybersecurity benchmarks: CyberGym, ExploitGym, and SEC-bench Pro. In CyberGym, it achieved an impressive score of 85.6%, surpassing its predecessor, GPT-5.5. In ExploitGym, which measures exploit generation capabilities, it scored 39.5%. Moreover, in SEC-bench Pro, focused on complex software vulnerability discovery, it achieved 69.8%.
This AI model is capable of navigating extensive codebases, identifying attack paths, assessing exploit potential, and generating precise patches, all within a streamlined automated workflow.
Enhanced Codex Security Plugin
Alongside GPT-5.5-Cyber, OpenAI has enhanced its Codex Security plugin, now featuring deep scanning capabilities and automated patch generation. Since its research preview release in March 2026, Codex Security has scanned over 30 million commits and processed over 70,000 manually verified fixes, automatically resolving more than 500,000 issues.
The plugin integrates into developer workflows, supporting SARIF exports and CodeQL queries, and contributes to existing vulnerability management systems by providing detailed severity reports and codebase-specific patches.
Collaborative Initiatives and Global Partnerships
OpenAI has also launched ‘Patch the Planet,’ a collaborative effort with Trail of Bits, HackerOne, and Calif, aimed at bridging the gap in open-source software vulnerability remediation. Over 30 open-source projects, including cURL, Go, Python, Sigstore, and pyca/cryptography, have joined this initiative.
A recent five-day sprint across participating projects resulted in hundreds of issues being identified and dozens of patches merged, with reusable workflows developed for future use. These projects benefit from access to ChatGPT Pro, Codex Security, and API credits.
GPT-5.5-Cyber is currently available exclusively to verified defenders through limited distribution. OpenAI has established Trusted Access partnerships with several countries, including Australia, Canada, France, and Japan, along with EU institutions. Pre-deployment testing was conducted in collaboration with the Center for AI Standards and Innovation (CAISI) and the Office of the National Cyber Director (ONCD) to align with AI security standards.
Future of Cyber Defense
This release marks a significant shift in cybersecurity paradigms, emphasizing the importance of patching vulnerabilities at scale. The Daybreak initiative integrates AI models, Codex Security workflows, and open-source collaborations, setting the stage for AI-driven remediation to become the cornerstone of cyber defense strategies.
