OpenAI officially launches a cybersecurity platform called Daybreak. This platform integrates the latest GPT-5.5 model with the Codex intelligent framework, shifting security defense from passive detection to automated response. Development teams can connect their code repositories to the system, allowing AI to automatically generate threat models, scan for high-risk vulnerabilities, and directly create fix codes (PRs) and regression tests within isolated work trees. Daybreak offers three model options, including the standard security-enabled GPT-5.5, as well as GPT-5.5 with Trusted Access for Cyber and GPT-5.5-Cyber, tailored for validation defense and specialized authorization workflows. The system can not only handle routine code vulnerability reviews but also directly compare against the latest CVEs (like CVE-2025-55182 demonstrated in the system), automatically query attack attempts in system logs, and instantly update firewall configurations. According to data disclosed by OpenAI, its previous-generation model GPT-5.4-Cyber has already helped fix over 3,000 high-risk vulnerabilities. For general users: Currently, the most advanced cybersecurity mode of Daybreak is not directly accessible. You’ll need to contact the OpenAI team to align with the most suitable security workflow model (click "Request a scan" on the official website to submit an application). For users with Codex access: Simply go to the Codex Security interface to use the basic features. Daybreak’s enhanced capabilities will be gradually rolled out in future iterations. Access here: https://(chatgpt.com)/codex/security (Make sure your Workspace settings have "Codex Cloud" permissions enabled first.)