OpenAI 正在扩展其 Daybreak 网络安全项目，为合作伙伴提供更高级的 AI 模型访问权限。该项目现在包含两个层级：Daybreak Blue 用于针对防御性安全量身定制的通用模型，Daybreak Red 用于专门的网络安全任务。新的 GPT-5.6-Cyber 模型基于 GPT-5.6 Sol 构建，旨在以更少的拒绝次数处理高风险任务，例如查找零日漏洞。此前，OpenAI 因担心其 Astra 模型可能被用于自主网络攻击而暂停了其开发。 AI
影响 增强了合作伙伴的防御性网络安全能力，同时也引发了对高级 AI 可能被滥用于进攻性网络任务的担忧。
排序理由 这是现有项目的扩展和专门模型的发布，而不是前沿模型的发布。
AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →
OpenAI's Daybreak program now has two tiers, with the higher one giving partners access to a more advanced cybersecurity model.
OpenAI is expanding access to cybersecurity AI models through the Daybreak Cyber Partner Program, with safeguards for the Astra model amid concerns about autonomous cyber tasks. Partners include Accenture, IBM, CrowdStrike, and others. Access remains controlled, with identity ver…
📰 OpenAI gives Daybreak partners access to a more powerful cybersecurity model OpenAI's Daybreak program now has two tiers, with the higher one giving partners access to a more advanced cybersecurity model. 📰 Source: Engadget - Technology News & Expert Reviews 🔗 Link: https://www…
OpenAI gives Daybreak partners access to a more powerful cybersecurity model https://www.engadget.com/2234335/openai-daybreak-model-less-likely-to-refuse-higher-risk-tasks/ # AI # Cybersecurity # Tech