OpenAI 推出 Daybreak 安全套件

openai launched daybreak, an ai initiative designed to detect and patch vulnerabilities before attackers can exploit them by utilizing the codex security agent to model attack paths and validate risks. the system integrates specialized models including gpt-5.5 with trusted access for cyber and the new gpt-5.5-cyber, which began rolling out last week. this release positions openai directly against anthropic's claude mythos and project glasswing, addressing openai's previous lack of a comparable security product. the company emphasized collaboration with industry and government partners as it prepares to deploy more advanced cyber-capable models.