OpenAI announced today the upgrade of its Daybreak cyber defense toolchain and the official release of the full GPT-5.5-Cyber model tailored for cybersecurity scenarios. On the CyberGym benchmark, which evaluates AI agents' ability to reproduce known vulnerabilities, GPT-5.5-Cyber achieved a single-model score of 85.6%, surpassing GPT-5.5's 81.8% and Anthropic Mythos 5's 83.8%. This demonstrates the model's enhanced capability in vulnerability identification and replication.
Shifting from Vulnerability Discovery to Automated Remediation
OpenAI noted that as AI significantly improves vulnerability discovery efficiency, the core bottleneck in cybersecurity is moving from "finding vulnerabilities" to "automatically fixing them." To address this, the company simultaneously upgraded the Codex Security plugin. This plugin enables developers to automatically analyze vulnerabilities and generate patches within the Codex environment. Since its preview release in March, Codex Security has scanned over 30 million code commits and confirmed fixes for 500,000 security defects. The new version also adds support for CodeQL query integration and SARIF standard file export, further expanding its applicability.
Open-Source Security Initiative and Ecosystem Collaboration
In addition, OpenAI, together with Trail of Bits and HackerOne, launched the "Patch the Planet" open-source security project. The initiative provides ChatGPT Pro subscriptions and API credits to more than 30 mainstream open-source projects, including cURL and Go. A team of security experts manually validates the quality of generated patches to ensure effectiveness and safety. On the Daybreak partner program front, security vendors such as Palo Alto Networks and Wiz have already integrated the relevant capabilities into their own product offerings.