微软发布首个网络安全模型与AI防御平台
TechCrunch
07-28 02:42
Ai Focus
微软发布网络安全专用 AI 模型和多代理防御平台,强化企业级安全产品布局。
Helpful
No.Help

微软在旧金山一场小型活动上发布首个网络安全专用模型 MAI-Cyber-1-Flash,同时推出新的 AI 安全平台 Perception。两项产品都面向企业安全团队,重点放在漏洞发现、修复和安全流程自动化。

模型接入漏洞检测

微软称,MAI-Cyber-1-Flash 专门用于在复杂代码库中寻找高难度漏洞,并已接入公司用于软件漏洞识别和修复的 MDASH 系统。按照微软的说法,该模型在一项业内常用的 AI 网络安全基准测试中,表现和成本效率都优于部分竞争产品。

微软 AI 负责人 Mustafa Suleyman 表示,这套组合已准备立即投入生产环境。微软在发布会上点名提到 Anthropic、Google 和 OpenAI,显示这次发布不仅是产品更新,也带有明显的竞争意味。

Perception部署多类代理

Perception 被设计成一套可部署多类智能体的安全平台,用来协助或自动完成多种安全工作流,包括识别漏洞、排序优先级和推动修复。该平台也可以与 MDASH 联动。

  • 红队代理负责模拟潜在攻击路径
  • 蓝队代理负责发现并分流现有漏洞
  • 绿队代理负责执行修复动作

微软安全业务高管 Hayete Gallot 表示,随着攻击者越来越多地使用 AI,企业防守方也需要用 AI 以相近的速度和规模应对攻击。

11月3日开放预览

Perception 负责人 Dave Weston 表示,过去需要多个安全岗位花费数小时完成的工作,现在可以在几分钟内完成。这套系统不仅能发现问题并确定优先级,还能给出检测方案、修复配置,甚至直接生成代码修复结果。

微软表示,这批新安全工具将于 11 月 3 日以预览形式推出。它们将进入一个竞争已经升温的市场。今年早些时候,Anthropic 推出安全平台 Mythos;OpenAI 也在 5 月通过 Day Break 项目发布了自己的安全方案。

Tip
$0
Like
0
Save
0
Views 595
CoinMeta reminds readers to view blockchain rationally, stay aware of risks, and beware of virtual token issuance and speculation. All content on this site represents market information or related viewpoints only and does not constitute any form of investment advice. If you find sensitive content, please click“Report”,and we will handle it promptly。
Submit
Comment 0
Hot
Latest
No comments yet. Be the first!
Related
Encore AI Raises $30 Million in Series A Funding
Encore AI has raised $30 million in Series A funding. The company focuses on training AI voice agents from customer calls, with most of its customers being financial institutions.
TechCrunch
·2026-07-29 22:53:54
215
Can brainwave data drive the implementation of physical AI?
Encord tests brainwave and electromyography data in an attempt to address the shortage of training data for robots.
TechCrunch
·2026-07-27 08:41:12
413
NVIDIA and 35 partners establish the Open AI Security Alliance
NVIDIA, together with more than 35 companies including Microsoft, Cisco, and IBM, has established the Open AI Security Alliance, focusing on the development of AI security and defense models.
Coinpedia
·2026-07-28 23:33:04
724
Ultraman on the Hugging Face incident: AI power should not be centralized
Altman stated that the Hugging Face incident highlights the security risks of AI and advocates for decentralizing AI power and capabilities to enhance the defense level of an open ecosystem.
Businessinsider
·2026-07-28 14:11:44
245
Nvidia, Microsoft, and IBM launch Open Security AI Alliance
NVIDIA, Microsoft, IBM, and others launched the Open Security AI Alliance to promote the development of open-source security tools and evaluation standards.
Decrypt
·2026-07-28 02:33:08
357