Microsoft 首个专门的网络安全模型 MAI-Cyber-1-Flash

The Hacker News

Microsoft Says New Cybersecurity AI Model Helps MDASH Score 95.95% at Half...

Microsoft launches MAI-Cyber-1-Flash inside MDASH, routing up to 90% of tasks to the model while GPT-5.4 handles the hardest 10%.

[!quote]+

公司表示，MDASH使用MAI-Cyber-1-Flash和GPT-5.4，在CyberGym中得分为95.95%。它还声称该配置的成本比目前最佳的MDASH组合（GPT-5.4、GPT-5.4 mini和GPT-5.3 Codex）低50%。访问仅限于通过 Azure AI Foundry 私有预览的获批 MDASH 客户。

MAI-Cyber-1-Flash 设计可处理高达 90% 的 MDASH 任务，GPT-5.4 则保留给最难的 10%。它仅在 MDASH 内部提供，不能作为独立的公共模型或通用应用程序接口。

主要得分属于MDASH与GPT-5.4并运行MAI-Cyber-1-Flash，而非新模型本身。CyberGym Level 1 是一个已知漏洞重现测试。它会给代理一个漏洞描述和相应的未打补丁源代码，然后检查是否能生成一个可工作的概念验证。它不测量盲目漏洞发现或生成补丁是否正确。

https://www.nytimes.com/2026/07/27/technology/microsoft-unveils-ai-cybersecurity-tools.html

TechCrunch – 27 Jul 26

Microsoft launches its first cybersecurity model, plus a new agentic...

Microsoft bolstered its AI cybersecurity offerings this week with the launch of its first AI security model and a new security platform.

Est. reading time: 3 minutes

https://www.axios.com/2026/07/27/microsoft-unveils-new-cyber-model-agentic-security-tools-to-fight-hackers