[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-microsoft-mai-cyber-1-flash-mdash":3,"news-related-82af5716-322e-46cf-9de5-b85e8cdd5712":33},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":20,"news_slug":26,"published_at":27,"created_at":28,"modified_at":29,"is_published":30,"publish_type":31,"image_url":13,"view_count":32},"82af5716-322e-46cf-9de5-b85e8cdd5712","微软首推安全专用模型 MAI-Cyber-1-Flash:小模型+多智能体,把漏洞挖掘成本砍半","7月27日,微软 AI 正式发布 MAI-Cyber-1-Flash,这是微软首个面向网络安全场景专门训练的语言模型,脱胎于 MAI-Thinking-1 家族,主打在大型代码库里识别高难度漏洞。模型并非以单点性能取胜,而是被嵌入到自家的多智能体漏洞识别与修复框架 MDASH 中:90% 的常规任务交给 MAI-Cyber-1-Flash 处理,剩下 10% 真正棘手的样本再交棒给 GPT-5.4。结果是,这套 MDASH + MAI-Cyber-1-Flash 的组合在 CyberGym 基准上拿到 95.95%,比 Mythos、Claude 等对手高出约 12 个百分点,同时相对纯 GPT-5.4 的多模型堆栈降低 50% 成本。\n\n更值得注意的是训练数据的护城河。微软每天承接超过 100 万亿条来自身份、终端、云、网络的安全信号,横跨 MSRC 漏洞库与 160 万企业客户的真实攻防数据,这种规模的历史经验几乎无法被复现。模型训练上,微软专门用安全优先的校准,红队 + 第三方独立评估,再叠加企业级 RBAC、租户隔离、加密、沙箱无外网等护栏。\n\n同一天,微软还推出了 agentic 安全系统 Perception,让多个 AI 智能体在 SOC 里持续监控、打补丁、闭环威胁,未来也将接入 MAI-Cyber-1-Flash。这标志着大厂打法正在切换:不再追求单一万能超大模型,而是训练小而专的小模型 + 多智能体编排 + 数据闭环,在垂直领域用极致性价比替代通用前沿模型。对安全行业来说,这是一次范式升级——攻防两端同时被 AI 加速,Defender 端的成本如果率先被打下来,主动防御就有可能从奢侈品变成标配。","https:\u002F\u002Fmicrosoft.ai\u002Fnews\u002Fintroducing-mai-cyber-1-flash-inside-mdash\u002F","9f1ed564-b7ac-4084-b1a8-4d129772db34",[10,14,17],{"id":11,"name":12,"slug":12,"description":13,"color":13},"e676a5cf-1f24-472f-a765-86fa21a1bc3c","ai-model",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",{"id":18,"name":19,"slug":19,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",[21],{"id":22,"lang":23,"title":24,"summary":25,"content":25},"8425cdca-58f2-4dbc-9030-4f05b741132a","en","Microsoft MAI-Cyber-1-Flash: security model halves bug-hunting cost","On July 27, Microsoft AI officially released MAI-Cyber-1-Flash — Microsoft's first language model specifically trained for cybersecurity scenarios, derived from the MAI-Thinking-1 family, and aimed at finding high-difficulty vulnerabilities in large codebases. The model isn't built to win on a single point metric; instead it's embedded inside Microsoft's in-house multi-agent vulnerability detection and remediation framework MDASH: 90% of routine tasks are handled by MAI-Cyber-1-Flash, and the remaining 10% truly tricky samples are then handed off to GPT-5.4. The result: this MDASH + MAI-Cyber-1-Flash combination scores 95.95% on the CyberGym benchmark, about 12 percentage points above Mythos, Claude and others, while reducing cost by 50% relative to a pure GPT-5.4 multi-model stack. The more noteworthy moat is the training data. Microsoft processes more than 100 trillion security signals per day from identity, endpoint, cloud, and network sources, spanning the MSRC vulnerability database and real-world attack\u002Fdefense data from 1.6 million enterprise customers — a scale of historical experience that's nearly impossible to replicate. The model is trained with a security-first calibration, red-teaming, and third-party independent evaluation, layered with enterprise-grade RBAC, tenant isolation, encryption, and sandboxed no-internet guardrails. On the same day, Microsoft also launched Perception, an agentic security system that lets multiple AI agents continuously monitor, patch, and close the loop on threats in the SOC — and Perception will also be wired into MAI-Cyber-1-Flash. This marks a shift in big-tech play: instead of chasing a single universal mega-model, the new approach is to train small, focused models + multi-agent orchestration + data feedback loops, and use extreme cost-effectiveness in vertical domains to replace general frontier models. For the security industry, this is a paradigm shift — offense and defense are both being accelerated by AI, and if the Defender side's cost is brought down first, proactive defense may turn from a luxury into a default.","microsoft-mai-cyber-1-flash-mdash","2026-07-28T01:00:00Z","2026-07-27T20:03:45.078519Z","2026-08-19T02:08:40.142862Z",true,"agent",239,{"items":34},[35,40,45,50,55,60],{"id":36,"title":37,"news_slug":38,"published_at":39},"a2e8ac5b-ca51-4ddb-88d4-54373d1f0774","SUNTA 用\"惊奇度\"切分视频预测:东京大学让模型在 250 步后仍不崩溃","sunta-surprise-chunking-video","2026-07-04T16:00:00+00:00",{"id":41,"title":42,"news_slug":43,"published_at":44},"774de6ac-98e1-4343-a67a-bfdc72d377bb","INFORMS 实证:AI 广告真实投放胜过设计师,18 个月后仍领先","informs-ai-ads-beat-human-designers-18-months","2026-08-22T14:00:00+00:00",{"id":46,"title":47,"news_slug":48,"published_at":49},"43eda321-b0b7-4df7-b20e-9758cbab42c9","记忆越完整,眼前题越做不对:MemTrapBench 把 LLM 长期记忆框架打回原形","memtrapbench-llm-memory-cognitive-traps","2026-08-22T04:00:00+00:00",{"id":51,"title":52,"news_slug":53,"published_at":54},"deac2d55-76a6-40d2-8ef7-36aed2ad0105","Linux 7.2 把 AI 拉进内核开发:Sashiko 让补丁数量翻倍,Torvalds 接受「新常态」","linux-7-2-sashiko-ai-kernel-review","2026-08-20T12:00:00+00:00",{"id":56,"title":57,"news_slug":58,"published_at":59},"fb1cbe25-8b85-41ec-b619-9a27b405ec34","AMD 收购 Taalas:把 AI 模型权重「刻进硅片」的推理新打法","amd-acquires-taalas-inference-chip","2026-08-19T01:00:00+00:00",{"id":61,"title":62,"news_slug":63,"published_at":64},"a91067a3-4fa4-4e88-a25a-18ba3bea21ea","Google 把\"加密推理\"摆上桌面：HEIR 编译器让预训练模型在密文上直接跑","google-heir-compiler-encrypted-ai-inference","2026-08-14T14:00:00+00:00"]