Microsoft 推出 MAI Cyber 2 Flash,与 GitHub 合作筛检开源安全报告
微软和 GitHub 拿 MAI Cyber 模型扫了六个开源项目,但每份报告都先过安全研究员的审核再发给维护者,99 份准确率有保障的报告而不是垃圾轰炸。做安全自动化、维护开源项目的人都该看看这套人机协作流程怎么设计。
Microsoft 的 Mustafa Suleyman 宣布发布 MAI Cyber 2 Flash,一个面向网络安全场景的模型。在此之前,MAI Cyber 前代模型已与 GitHub 合作,扫描了六个广泛使用的开源项目,寻找潜在安全漏洞。
这一项目的关键设计是人工审核前置:AI 生成的每份漏洞报告先由安全研究人员逐条核查,确认后才转交维护者。六个项目中,维护者最终收到 99 份经过核查的私密报告,而不是被大量未经筛选的 AI 生成 bug 报告淹没,部分修复已在推进。
开源维护者长期面临 AI 生成的低质量漏洞报告泛滥的问题,这类噪音挤占了本应用于修复漏洞的时间。MAI Cyber 2 Flash 与 GitHub 的这一流程将人工审核作为交付环节,可能成为 AI 安全挖掘工具与开源社区协作的参照模式。
Today we announced MAI Cyber 2 Flash, and here’s some of the work that came before it… Open source runs the world’s software. Its maintainers are now flooded with AI-generated bug reports, and sorting through them takes time away from fixing. We tried something different with GitHub. Our earlier MAI Cyber model scanned six widely used open source projects. Security researchers checked each report before any maintainer saw it. Maintainers got 99 checked, private reports instead of a flood, and they’re already shipping fixes. That’s what good AI security should look like: AI findings checked by people before they reach maintainers.📊 1 ⚡
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include MAI Cyber, GitHub, open source, AI safety.
Source attribution: curated from Mustafa Suleyman (@mustafasuleyman).