OpenAI 发布 GPT-5.6-Cyber 模型，助力防御者提前发现漏洞
原标题：OpenAI launches GPT-5.6-Cyber to help defenders find vulnerabilities before attackers do
AI 摘要
OpenAI 扩展了其 Daybreak 网络安全计划，新增两个访问层级和专门模型 GPT-5.6-Cyber，旨在帮助防御者提前发现漏洞。该模型基于 GPT-5.6 Sol 训练，能回答 95% 的敏感安全查询，并已发现 Chrome 的 V8 引擎中的两个未知漏洞。OpenAI 表示，该模型在网络安全能力上被评为“高”，但未达到“严重”阈值。
正文节选
OpenAI launches GPT-5.6-Cyber to help defenders find vulnerabilities before attackers do Key Points - OpenAI is expanding its Daybreak cybersecurity program with two new access tiers and a dedicated AI model called GPT-5.6-Cyber, designed to help security professionals identify vulnerabilities and develop exploits at an early stage. - The program is split into two tracks: Daybreak Blue focuses on defensive tasks such as malware analysis, while Daybreak Red is geared toward offensive security res