[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-anthropic-alibaba-claude-distillation-attack-zh":3,"article-related-anthropic-alibaba-claude-distillation-attack-zh":32,"series-industry-218528dc-b109-45ba-b6f2-6b8181b9c84d":79},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":24,"views":28,"created_at":29,"published_at":30,"topic_cluster_id":31},"218528dc-b109-45ba-b6f2-6b8181b9c84d","anthropic-alibaba-claude-distillation-attack-zh","Anthropic 指控 Alibaba 大量蒸餾 Claude","\u003Cp data-speakable=\"summary\">\u003Ca href=\"\u002Ftag\u002Fanthropic\">Anthropic\u003C\u002Fa> 指控 Alibaba 用假帳號與數千萬次 \u003Ca href=\"\u002Ftag\u002Fclaude\">Claude\u003C\u002Fa> 呼叫，蒐集輸出訓練競品模型。\u003C\u002Fp>\u003Cp>這件事不是單純的政策違規。它碰到的是 AI 產業最怕的事：你的模型回答，可能變成別人的訓練資料。\u003C\u002Fp>\u003Cp>Anthropic 說，規模大到不像零星測試。25,000 個假帳號，外加 2,880 萬次 Claude 互動，聽起來就是系統化操作。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>數字\u003C\u002Fth>\u003Cth>意義\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>假帳號數\u003C\u002Ftd>\u003Ctd>25,000\u003C\u002Ftd>\u003Ctd>像是有組織的批量操作\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Claude 互動次數\u003C\u002Ftd>\u003Ctd>28.8 million\u003C\u002Ftd>\u003Ctd>代表資料蒐集規模很大\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>指向對象\u003C\u002Ftd>\u003Ctd>Claude\u003C\u002Ftd>\u003Ctd>熱門前沿 LLM，輸出很有訓練價值\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>Anthropic 指控了什麼\u003C\u002Fh2>\u003Cp>Anthropic 的說法很直接。有人用大量假帳號去呼叫 \u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\u002Fclaude\" target=\"_blank\" rel=\"noopener\">Claude\u003C\u002Fa>，再把回應拿去訓練別的模型。講白了，就是把別人的 \u003Ca href=\"\u002Ftag\u002Fapi\">API\u003C\u002Fa> 當資料工廠。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782583382703-xndr.png\" alt=\"Anthropic 指控 Alibaba 大量蒸餾 Claude\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這裡的重點不是「有沒有用蒸餾」。蒸餾本來就是 AI 開發常見手法。重點是，你有沒有授權、怎麼抓資料、抓到什麼程度。\u003C\u002Fp>\u003Cp>如果是未經同意的大量擷取，那就不是一般工程技巧。這會碰到服務條款、資安控管，甚至法律問題。\u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\u002Fnews\" target=\"_blank\" rel=\"noopener\">Anthropic Newsroom\u003C\u002Fa> 近年一直強調使用限制與監控，這次等於是自己把槍口對準濫用案例。\u003C\u002Fp>\u003Cblockquote>“We are seeing a rising tide of AI misuse.” — Dario Amodei, Anthropic CEO\u003C\u002Fblockquote>\u003Cul>\u003Cli>25,000 個帳號，像是有工具在批次輪替。\u003C\u002Fli>\u003Cli>2,880 萬次互動，足以形成大規模語料。\u003C\u002Fli>\u003Cli>Claude 的輸出品質高，特別適合拿來做 instruction tuning。\u003C\u002Fli>\u003Cli>API 濫用會把產品使用，變成資料收割。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>為什麼這組數字很刺眼\u003C\u002Fh2>\u003Cp>25,000 個假帳號，不像臨時起意。這種數字通常代表自動化、\u003Ca href=\"\u002Fnews\u002Fai-payment-bots-strict-limits-web3-zh\">代理\u003C\u002Fa>池，還有一套躲 rate limit 的流程。\u003C\u002Fp>\u003Cp>2,880 萬次互動也不是小事。這種量級，已經不是「試看看」而已，而是長時間、持續性的蒐集。\u003C\u002Fp>\u003Cp>這也是 AI 業界現在最頭痛的地方。模型輸出很有用，但也很容易被拿去當訓練資料。你提供越好的 API，別人越想把你的回答變成自己的模型能力。\u003C\u002Fp>\u003Cul>\u003Cli>25,000 帳號可支援分散式查詢。\u003C\u002Fli>\u003Cli>2,880 萬次呼叫可快速堆出訓練集。\u003C\u002Fli>\u003Cli>高品質輸出越多，蒸餾價值越高。\u003C\u002Fli>\u003Cli>規模一大，問題就從產品濫用變成競爭風險。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>說真的，這就是 API 時代的陰影面。你賣的是智慧回答，對方買到的可能是資料來源。\u003C\u002Fp>\u003Ch2>Alibaba 牽涉到哪裡\u003C\u002Fh2>\u003Cp>這次被點名的是 \u003Ca href=\"https:\u002F\u002Fwww.alibabagroup.com\" target=\"_blank\" rel=\"noopener\">Alibaba Group\u003C\u002Fa>。但真正該查的，不只是公司名稱。還要看是內部團隊、外包商，還是某個合作單位在操作。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782583380332-4yax.png\" alt=\"Anthropic 指控 Alibaba 大量蒸餾 Claude\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>大公司常常有很多層。研究團隊、產品團隊、雲端部門、合作夥伴，全部都可能碰到\u003Ca href=\"\u002Fnews\u002Fai-web3-equal-rules-not-special-gates-zh\">同一套\u003C\u002Fa>基礎設施。責任歸屬沒釐清前，外界很難只看一個名字就下結論。\u003C\u002Fp>\u003Cp>Alibaba 自己也有 AI 佈局，像 \u003Ca href=\"https:\u002F\u002Fwww.alibabacloud.com\u002Fproduct\u002Fqwen\" target=\"_blank\" rel=\"noopener\">Qwen\u003C\u002Fa>。所以這個指控一旦成立，外界一定會追問：資料從哪來，誰拿去訓練，內控在哪裡失靈。\u003C\u002Fp>\u003Cul>\u003Cli>誰建立了 25,000 個帳號？\u003C\u002Fli>\u003Cli>誰付了 2,880 萬次呼叫的成本？\u003C\u002Fli>\u003Cli>資料最後進了訓練集，還是只做評估？\u003C\u002Fli>\u003Cli>有沒有違反服務條款或安全政策？\u003C\u002Fli>\u003C\u002Ful>\u003Cp>這些問題比網路吵架重要多了。AI 爭議最後都會回到證據。\u003C\u002Fp>\u003Ch2>拿其他 AI 濫用案例來看\u003C\u002Fh2>\u003Cp>這種事不是第一次發生。\u003Ca href=\"https:\u002F\u002Fopenai.com\" target=\"_blank\" rel=\"noopener\">OpenAI\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fdeepmind.google\" target=\"_blank\" rel=\"noopener\">Google DeepMind\u003C\u002Fa>、Anthropic 這類公司，早就面對 prompt scraping、帳號濫用、API 批量查詢。\u003C\u002Fp>\u003Cp>問題一直都在，只是以前沒有這麼大聲。當模型越來越好用，輸出就越值錢。有人會想辦法把輸出壓縮、重組，再拿去訓練自己的系統。\u003C\u002Fp>\u003Cp>這不是理論題。這是商業現實。你做出一個能穩定回應的 \u003Ca href=\"\u002Ftag\u002Fllm\">LLM\u003C\u002Fa>，等於也做出一個很吸引人的資料來源。\u003C\u002Fp>\u003Cul>\u003Cli>API 濫用是所有大模型供應商都會遇到的事。\u003C\u002Fli>\u003Cli>蒸餾本身不違法，前提是有授權。\u003C\u002Fli>\u003Cli>未授權擷取會變成資安與法務問題。\u003C\u002Fli>\u003Cli>規模越大，越容易踩到紅線。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>對開發者來說，這代表監控不能只看 latency 和 error rate。還要看帳號行為、查詢模式、異常流量。\u003C\u002Fp>\u003Cp>對產品團隊來說，這也很現實。你想開放 API，就得接受有人會想辦法把你的模型吃乾抹淨。\u003C\u002Fp>\u003Ch2>這件事對產業的意思\u003C\u002Fh2>\u003Cp>AI 產業現在很像一場拉鋸。模型公司想更\u003Ca href=\"\u002Fnews\u002Fai-agents-web3-strict-controls-not-hype-zh\">嚴格控管\u003C\u002Fa>。使用者和競爭者則一直找漏洞。兩邊都知道，輸出本身就是價值。\u003C\u002Fp>\u003Cp>如果 Anthropic 的數字站得住腳，之後大家會更重視驗證、登入、風控與稽核。這類成本會上升，而且不會只出現在 Anthropic。\u003C\u002Fp>\u003Cp>我覺得這才是重點。不是誰被罵得比較兇，而是整個產業會不會開始把「防資料外流」當成基本配備。\u003C\u002Fp>\u003Cul>\u003Cli>模型供應商會加強異常偵測。\u003C\u002Fli>\u003Cli>企業客戶會更在意資料來源。\u003C\u002Fli>\u003Cli>API 價格可能反映更多風控成本。\u003C\u002Fli>\u003Cli>訓練資料治理會變成採購條件。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>這也會影響 SaaS、雲端服務，還有所有把 LLM 接進產品的團隊。你以為你在用 API，其實你也在參與一場資料攻防戰。\u003C\u002Fp>\u003Ch2>接下來該看什麼\u003C\u002Fh2>\u003Cp>下一步要看 Anthropic 會不會拿出更多證據。像是登入紀錄、IP 分布、帳號關聯，或是流量模式。如果證據夠硬，這案子可能會變成業界參考案例。\u003C\u002Fp>\u003Cp>如果指控最後站不穩，市場也不會當沒事。因為所有模型公司都會更緊張，接著把審核、驗證、封鎖機制再往上加。\u003C\u002Fp>\u003Cp>講白了，這場爭議提醒大家一件事：AI 競爭不只看 \u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa>。還要看誰能守住資料邊界。開發者接下來最好假設一件事，模型輸出一定會有人想拿去再訓練。\u003C\u002Fp>\u003Cp>你如果在做 AI 產品，現在就該檢查帳號風控、API 配額、異常查詢偵測。這種事拖下去，通常只會更貴。\u003C\u002Fp>","Anthropic 指控 Alibaba 透過 25,000 個假帳號與 2,880 萬次 Claude 呼叫，蒐集輸出訓練競品模型。這起爭議把模型蒸餾、API 濫用與資料治理，直接拉到企業級規模。","zhuanlan.zhihu.com","https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2053757165005967725",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782583382703-xndr.png","industry","zh","25bce581-e9e6-4070-9665-98eb144c6f97",[17,18,19,20,21,22,23],"Anthropic","Alibaba","Claude","模型蒸餾","API 濫用","AI 資安","Qwen",[25,26,27],"Anthropic 指控 Alibaba 用 25,000 個假帳號與 2,880 萬次 Claude 互動蒐集資料。","這起事件把模型蒸餾問題，拉到企業級規模與法務層面。","對 AI 開發者來說，帳號風控與異常查詢監控已經是基本配備。",0,"2026-06-27T18:02:38.014434+00:00","2026-06-27T18:02:38.005+00:00","2022f530-e831-4f6f-8dd0-9b7b7f165398",{"tags":33,"relatedLang":38,"relatedPosts":42},[34,36],{"name":17,"slug":35},"anthropic",{"name":19,"slug":37},"claude",{"id":15,"slug":39,"title":40,"language":41},"anthropic-alibaba-claude-distillation-attack-en","Anthropic Accuses Alibaba of Massive Claude Distillation","en",[43,49,55,61,67,73],{"id":44,"slug":45,"title":46,"cover_image":47,"image_url":47,"created_at":48,"category":13},"f3ee7f58-9ef7-4846-95c3-839462c0347d","openclaw-openai-realtime-paid-api-not-subscription-perk-zh","OpenClaw 應把 OpenAI Realtime 當付費 API，而不是…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782674270289-mop4.png","2026-06-28T19:17:24.429354+00:00",{"id":50,"slug":51,"title":52,"cover_image":53,"image_url":53,"created_at":54,"category":13},"b90d3831-5109-404e-89a5-50c4890910ed","krea-2-two-second-image-generation-teams-zh","Krea 2 的 2 秒生成，適合團隊部署嗎","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782673366702-fv0o.png","2026-06-28T19:02:22.494136+00:00",{"id":56,"slug":57,"title":58,"cover_image":59,"image_url":59,"created_at":60,"category":13},"576f1de0-bbf9-4a91-96bf-a1bf6ff4c67c","us-model-curbs-security-deals-not-bans-zh","美國應以安全協議解除模型管制，而非一刀切禁令","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782658969312-tf30.png","2026-06-28T15:02:19.927898+00:00",{"id":62,"slug":63,"title":64,"cover_image":65,"image_url":65,"created_at":66,"category":13},"c1d71ae5-dabd-4778-8326-7645316004c2","meta-replacing-moderators-with-ai-to-cut-costs-zh","Meta 用 AI 取代審核員，省錢先上","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782653576451-arn6.png","2026-06-28T13:32:29.737246+00:00",{"id":68,"slug":69,"title":70,"cover_image":71,"image_url":71,"created_at":72,"category":13},"08c94bd8-e6b6-4328-82ff-bee0a7cef126","meta-ai-moderation-push-is-the-wrong-tradeoff-zh","Meta 把 AI 用在內容審核上，這筆交換不划算","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782652669314-in2k.png","2026-06-28T13:17:21.733509+00:00",{"id":74,"slug":75,"title":76,"cover_image":77,"image_url":77,"created_at":78,"category":13},"6ad43bed-fc6b-4bc6-a728-38362a29ffec","meta-ai-content-moderation-human-reviews-zh","Meta 內容審核轉向 AI 的 5 個關鍵","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782651773409-llaq.png","2026-06-28T13:02:22.855907+00:00",[80,85,90,95,100,105,110,115,120,125],{"id":81,"slug":82,"title":83,"created_at":84},"ee073da7-28b3-4752-a319-5a501459fb87","ai-in-2026-what-actually-matters-now-zh","2026 AI 真正重要的事","2026-03-26T07:09:12.008134+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"83bd1795-8548-44c9-9a7e-de50a0923f71","trump-ai-framework-power-speech-state-preemption-zh","川普 AI 框架瞄準電力、言論與州權","2026-03-26T07:12:18.695466+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"ea6be18b-c903-4e54-97b7-5f7447a612e0","nvidia-gtc-2026-big-ai-announcements-zh","NVIDIA GTC 2026 重點拆解","2026-03-26T07:14:26.62638+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"4bcec76f-4c36-4daa-909f-54cd702f7c93","claude-users-spreading-out-and-getting-better-zh","Claude 用戶更分散，也更會用","2026-03-26T07:22:52.325888+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"bd903b15-2473-4178-9789-b7557816e535","openclaw-raises-hard-question-for-ai-models-zh","OpenClaw 逼問 AI 模型價值","2026-03-26T07:24:54.707486+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"eeac6b9e-ad9d-4831-8eec-8bba3f9bca6a","gap-google-gemini-checkout-fashion-search-zh","Gap 把結帳搬進 Gemini","2026-03-26T07:28:23.937768+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"0740e53f-605d-4d57-8601-c10beb126f3c","google-pushes-gemini-transition-to-march-2026-zh","Google 把 Gemini 轉換延到 2026 年 3…","2026-03-26T07:30:12.825269+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"e660d801-2421-4529-8fa9-86b82b066990","metas-llama-4-benchmark-scandal-gets-worse-zh","Meta Llama 4 分數風波又擴大","2026-03-26T07:34:21.156421+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"183f9e7c-e143-40bb-a6d5-67ba84a3a8bc","accenture-mistral-ai-sovereign-enterprise-deal-zh","Accenture 攜手 Mistral AI 賣主權 AI","2026-03-26T07:38:14.818906+00:00",{"id":126,"slug":127,"title":128,"created_at":129},"191d9b1b-768a-478c-978c-dd7431a38149","mistral-ai-faces-its-hardest-year-yet-zh","Mistral AI 迎來最硬的一年","2026-03-26T07:40:23.716374+00:00"]