2026
84 篇文章Claude 免費用戶的資料開關:訓練授權、五年保留期與產品團隊要留意的界線
Anthropic 更新消費者條款,讓 Claude Free、Pro、Max 用戶自行決定是否用對話資料訓練模型,並把保留期從 30 天延長到五年。
閱讀文章 ↗當 AI 走進國防與公部門:Anthropic 顧問團對產品團隊的三個現實提醒
Anthropic 成立國家安全與公部門顧問團,本文拆解這對做 AI 產品的人在合規、部署與標準上的實際影響。
閱讀文章 ↗把 AI 帶進教室前,先想清楚要教什麼:Anthropic 高教顧問團與 AI Fluency 課程的啟示
Anthropic 成立高教顧問團並推出三門 AI Fluency 課程,為教育機構提供可改編的實用框架,而非抽象原則。
閱讀文章 ↗當核安遇上分類器:Anthropic 與 NNSA 的合作,對開發者意味著什麼
Anthropic 與美國 NNSA 共同開發核相關內容分類器,初步測試準確率 96%,並已部署於 Claude 流量。
閱讀文章 ↗當 Claude Code 被拿來勒索:產品開發者該從 Anthropic 8 月威脅報告讀到什麼
Anthropic 8 月威脅報告揭露 Claude Code 被用於自動化勒索、北韓假員工與 AI 生成勒索軟體,本文拆解對產品開發者的三個訊號。
閱讀文章 ↗Anthropic 與蓋茲基金會合作:AI 如何進入市場不願觸及的領域
Anthropic 與蓋茲基金會投入 2 億美元,用 Claude 支援全球健康、教育與經濟流動性計畫,並釋出公共工具。
閱讀文章 ↗AI 代理如何把網路間諜活動從「人為指揮」變成「自主執行」
Anthropic 揭露首宗大規模 AI 主導的網路間諜活動,Claude Code 被濫用執行 80-90% 攻擊流程,開發者需重新思考代理式 AI 的防禦設計。
閱讀文章 ↗從 Claude 濫用案例看 AI 安全:產品開發者該注意的三個訊號
Anthropic 揭露 Claude 被用於影響力操作、憑證填充、招募詐騙與惡意軟體開發的案例。這對正在建構 AI 產品的開發者,意味著安全設計不能只靠事後偵測,而要從產品初期就納入濫用情境的考量。
閱讀文章 ↗蒸餾攻擊不是理論:Anthropic 揭露 DeepSeek、Moonshot、MiniMax 的 1,600 萬次提取
Anthropic 在 2026 年 2 月揭露三家 AI 實驗室透過 24,000 個詐騙帳號,對 Claude 發動工業級蒸餾攻擊。本文從產品建構者角度,解析攻擊手法、偵測機制,以及這對 API 安全與出口管制的啟示。
閱讀文章 ↗Claude 十一天形式化費馬最後定理:1,300 萬行 Lean 完整證明
2026 年 9 月,Anthropic 宣布 Claude 代理群在 11 天內完成費馬最後定理的完整 Lean 形式化:1,300 多萬行程式碼、逾 3 萬條定理,由數學家 Kevin Buzzard 親自驗證。
閱讀文章 ↗從 Claude 越獄事件看 AI 安全:產品開發者的實用啟示
Anthropic 公開了 Claude 模型未經授權存取真實系統的事件,並分享了安全與對齊的改進措施。本文為產品開發者解析這些事件背後的教訓,以及如何在 AI 工具開發中落實更穩健的安全實踐。
閱讀文章 ↗Anthropic 包下 SpaceX Colossus 1:向對手租整座資料中心的多源算力策略
2026 年 5 月,Anthropic 取得 Memphis Colossus 1 全部容量——超過 22 萬顆 GPU、300 多 MW,用於改善 Claude Pro 與 Max 的容量,並表態合作開發 GW 級軌道算力。本文解析這筆「向對手租機房」的交易,與 Google、Amazon 交易串成的多源算力佈局。
閱讀文章 ↗當 AI 測試環境意外連上真實網路:Anthropic 三起事件的檢討
Anthropic 在回顧網路安全評估時,發現 Claude 模型因環境設定錯誤而意外存取真實系統。本文整理事件經過、原因與後續改進,並探討對 AI 評估與產品建構者的啟示。
閱讀文章 ↗Anthropic 與 OpenAI 的州級 AI 安全法案之戰:棘輪式升級 vs 反向聯邦主義
美國國會立法停擺下,AI 實驗室轉戰州議會。Anthropic 力推各州逐級加嚴(伊利諾州已簽署全美首見的第三方稽核法),OpenAI 則主打「同一套法案複製到各州」;兩家在伊利諾州 SB 3444 免責條款上正面交鋒。本文整理雙方策略與各州進展。
閱讀文章 ↗Fable 5.1 與 Mythos 5.1 同模型:快取讀取降 75%
Anthropic 同日發布 Fable 5.1 與 Mythos 5.1:兩者是同一個模型、不同護欄等級,Fable 對一般市場開放,Mythos 走受信存取計劃、專為網路安全與生命科學研究設計,快取讀取降 75%、資安誤攔每 session 減 60%。
閱讀文章 ↗Enterprise Frontier Safeguards:把監控資料留在客戶手上,同時守住前沿模型安全
Anthropic 與超過一百家企業客戶共同設計 Enterprise Frontier Safeguards,結合零資料保留與跨時段、跨帳號的濫用偵測。本文從產品建構者的角度拆解這套架構如何回應監管、資料主權與人為審查的實際需求。
閱讀文章 ↗Claude for Teachers:把教學標準和備課時間,還給老師
Anthropic 推出 Claude for Teachers,為美國 K-12 教師提供免費的 Claude 高階功能、教學技能庫,並串接全美 50 州的學術標準與實證課程。本文整理產品重點、資料隱私設計,以及對教育科技建構者的啟示。
閱讀文章 ↗Anthropic 擴大科學家支援:10,000 個免費與折扣 Claude 訂閱席位
Anthropic 宣布開放 10,000 個科學家席位,提供一年免費或折扣的 Claude 訂閱,並擴大 AI for Science 計畫至生物學以外的領域。
閱讀文章 ↗Model Hardware Standard 研究預覽:讓 AI Agent 安全操作實驗室與工廠設備
Anthropic 公開 Model Hardware Standard(MHS)研究預覽,以標準化驅動程式讓 AI agent 平行操作顯微鏡、液體處理器與機械臂,並分享早期合作夥伴在生技、量子運算與製造領域的測試結果。
閱讀文章 ↗100 萬筆對話實證:57% 的 Claude 使用屬增強而非自動化
Anthropic 分析 Claude 3.7 Sonnet 上線後 11 天內 100 萬筆匿名對話:增強型使用佔 57% 不變、學習型互動升至 28%,編碼教育科學成長最快,630 類使用分類資料集已公開於 Hugging Face。
閱讀文章 ↗用 Claude 直接查詢 Anthropic Economic Index:AI 對工作的影響,現在可以問了
Anthropic 推出 Economic Index connector,讓用戶在 claude.ai 直接以自然語言查詢 AI 對職業與任務的影響數據。本文介紹功能、使用方式與限制。
閱讀文章 ↗Anthropic 撥款五百萬美元,資助 AI 對幸福感影響的獨立評估
Anthropic 推出五百萬美元資助計劃,支持獨立研究 AI 對用戶幸福感的影響,並公開評估指引,涵蓋多輪對話、臨床專家參與、評分驗證、申請時程與成果開源要求。
閱讀文章 ↗Claude Sonnet 5:把 Opus 級能力帶到 Sonnet 價格帶
Anthropic 推出 Claude Sonnet 5,宣稱是迄今最具 agentic 能力的 Sonnet 模型,效能接近 Opus 4.8,但價格更低。本文整理其能力、定價、安全評估與開發者回饋,並討論對產品建構者的意義。
閱讀文章 ↗Dario Amodei 表態:Anthropic 從未主張禁用開放權重模型
美國官員傳出考慮禁用中國開放權重模型、科技業連署力挺開放權重之際,Amodei 於 7 月 27 日撰文澄清 Anthropic 從未主張禁令,改提出晶片出口管制、打擊工業級蒸餾與強制安全測試三項主張。
閱讀文章 ↗Claude Mythos 60 小時改寫 HAWK 攻擊
Anthropic 前沿紅隊報告:Claude Mythos Preview 用 60 小時把 HAWK-256 攻擊成本從 2^64 降到 2^38,並以 Möbius Bridge 技術把 7 輪 AES-128 攻擊加速 200 至 800 倍。
閱讀文章 ↗Anthropic 與 Cognizant 擴大合作:把 Claude 帶進企業的關鍵在於「領域知識」
Anthropic 與 Cognizant 擴大合作,後者將 Claude 嵌入自家平台與客戶系統,並培訓超過 3 萬名員工。本文從產品思維角度,拆解這次合作對企業採用 AI 的啟示。
閱讀文章 ↗Anthropic 15 億美元圖書和解案獲批:每本書 3,000 美元
聯邦法官 Araceli Martinez-Olguin 最終批准 Anthropic 15 億美元圖書侵權和解,約 50 萬本書、每本 3,000 美元;但 AI 訓練資料的版權之戰才剛進入下一輪。
閱讀文章 ↗Claude Fable 推翻雅可比猜想:數學界震動的一週
Anthropic 研究員 Levent Alpöge 與 Claude Fable 5 在世界盃決賽夜提出雅可比猜想反例,隔天完成 Lean 驗證、陶哲軒撰文解析,AI 正在改寫數學研究的驗證流程。
閱讀文章 ↗Anthropic 罕見疾病研究補助:$50,000 Claude credits 與兩軌設計
Anthropic 在 AI for Science 計畫下開出罕見遺傳疾病主題徵案:六個月內最高 $50,000 的 Claude credits,分基礎科學與早期 biotech 兩軌。本文拆解兩軌的機制差異、Monarch 的資料基礎建設、三個既有受補助案例,以及 Anthropic 自己畫出的能力邊界。
閱讀文章 ↗Anthropic 的 Safeguards 團隊如何為 Claude 建立多層防護
Anthropic 公開 Safeguards 團隊的運作方式:從政策制定、模型訓練、測試評估到即時偵測,多層次確保 Claude 安全可靠。本文為產品開發者拆解這套防護架構的實際做法與挑戰。
閱讀文章 ↗Anthropic 洽談三星代工自研晶片:多源算力的第四條線
The Information 報導 Anthropic 正與三星洽談代工自研 AI 晶片,用途與規格皆未定;在 Google TPU、Amazon Trainium 與 Nvidia GPU 之外,多源硬體策略可能再添一條自己的供應線。
閱讀文章 ↗Fable 5 回歸:新分類器擋下 99% 越獄手法
Fable 5 因美國出口管制暫停後於 7 月 1 日恢復;新分類器擋下 99% 越獄手法,四準則嚴重度框架與 Amazon 等共同草擬中。
閱讀文章 ↗AWS 投入 10 億美元成立 FDE 組織,把工程師送進客戶現場
AWS 宣布投入 10 億美元成立專責 FDE 組織,把前瞻 AI 工程師派駐客戶內部部署代理式系統,跟進 OpenAI 與 Anthropic 的企業服務路線,但選擇不與私募股權合資。
閱讀文章 ↗付費消費者轉向 Claude:ChatGPT 護城河出現裂縫
信用卡消費數據、學習平台搜尋行為與第三方報告同時指向一件事:付費個人用戶正加速轉向 Claude,2026 年初以來人數與營收成長約 75%。ChatGPT 在各項指標仍全面領先,但優勢不再牢不可破。
閱讀文章 ↗Anthropic 的 Claude Corps:用一年時間,把 AI 技能帶進非營利組織
Anthropic 推出 Claude Corps 獎學金計畫,培訓 1,000 名職涯初期人才,在美國非營利組織全職工作一年,並投入 1.5 億美元。本文整理計畫細節、合作夥伴與背後的勞動市場思考。
閱讀文章 ↗Claude 可能要看你的證件:Anthropic 驗證政策與生物特徵爭議
TechCrunch 報導,Anthropic 更新隱私政策(7 月 8 日生效),特定情況下可要求 Claude 使用者上傳護照或駕照加自拍,並產生臉部幾何模板驗證身分,引發生物特徵資料保留與管轄權疑慮。
閱讀文章 ↗諾貝爾獎得主 John Jumper 宣布離開 DeepMind 轉戰 Anthropic
2024 年諾貝爾化學獎共同得主 John Jumper 宣布結束近九年的 DeepMind 生涯轉投 Anthropic,同週 Transformer 論文共同作者 Noam Shazeer 也離開 Google 前往 OpenAI,前沿實驗室的人才爭奪白熱化。
閱讀文章 ↗報導指出 SK Telecom 是 Claude Fable 5 出口管制事件的導火線
2026 年 6 月 19 日,報導指出 SK Telecom 是美國 6 月暫停 Claude Fable 5 出口管制事件的真正觸發者。本文回顧 6 月 12 日對 Fable 5 與 Mythos 5 的管制與停權,並談對 model 供應風險的啟示。
閱讀文章 ↗Anthropic Project Fetch 第二階段:Opus 4.7 操作機器狗比人快 20 倍
Anthropic 於 2026 年 6 月 18 日發布 Project Fetch 第二階段:Claude Opus 4.7 全自主操作機器狗,比最快人類團隊快約 20 倍、程式碼少十倍,卻仍推不動那顆海灘球。本文拆解數據、方法與社群爭議。
閱讀文章 ↗Anthropic 被要求停用 Fable 5 與 Mythos 5:一次關於監管與 AI 安全的警鐘
美國政府以國家安全為由,要求 Anthropic 暫停所有用戶對 Fable 5 與 Mythos 5 的存取。Anthropic 公開回應,質疑此決定的技術依據,並指出此舉可能開創監管先例。本文整理事件始末,並探討對 AI 產品開發者的啟示。
閱讀文章 ↗美國人點睇 AI?Anthropic 首份 Public Record 調查的 5 個關鍵發現
Anthropic 發布首份 Public Record 調查,訪問近 52,000 名美國人。結果顯示:疾病治療是最大期望、失業是最普遍擔憂、跨黨派支持監管,但僅 15% 信任 AI 公司。本文整理對產品建構者的啟示。
閱讀文章 ↗Anthropic 與 TCS 結盟:把 Claude 交付規模化的通路之戰
2026 年 6 月 11 日,Anthropic 與 TCS 宣布全球合作:TCS 設專責事業單位部署 Claude,5 萬名員工率先導入,並切入金融、醫療、電信與航空垂直。本文解析企業 AI 通路之戰,與印度 IT 服務業的轉型賭注。
閱讀文章 ↗AI 令網路攻擊更危險,但現有框架可能看不見
Anthropic 分析 832 個因惡意網路活動被停用的帳戶,發現 AI 正被用於攻擊鏈的後期階段,令攻擊更自動化,而 MITRE ATT&CK 框架未能完全捕捉這些新行為。
閱讀文章 ↗Anthropic 開設米蘭辦公室:意大利企業與開發者的 AI 落地新據點
Anthropic 於 2026 年 5 月宣布在米蘭開設第六個歐洲辦公室,聚焦意大利企業、研究與開發者社群,並已與多家大型企業合作部署 Claude。本文整理其具體案例與背後策略。
閱讀文章 ↗Claude Opus 4.8 發布:Dynamic Workflows 與更誠實的程式碼模型
Anthropic 在 41 天後推出 Claude Opus 4.8:價格不變,新增可協調上百個子代理的 Dynamic Workflows 與 Effort 控制,SWE-bench Pro 升至 69.2%,Mythos 預告數週內開放。
閱讀文章 ↗Anthropic 在首爾設辦公室,找來 KiYoung Choi 領導韓國市場
韓國是 Claude.ai 全球最活躍的市場之一,使用率比人口比例高出 3.5 倍。Anthropic 任命 KiYoung Choi 為韓國代表董事,並將在首爾開設辦公室,深入經營企業與開發者生態。
閱讀文章 ↗Anthropic 逾 300 億美元新輪進入收官:9,000 億估值將改寫排名
2026 年 5 月 23 日,Reuters 報導 Anthropic 可望最快下週完成逾 300 億美元新輪募資,估值超過 9,000 億美元,FT 稱條款已敲定。本文解析資金規模與後續觀察指標。
閱讀文章 ↗Karpathy 宣布加入 Anthropic:pre-training 人才流向的風向球
2026 年 5 月 19 日,OpenAI 共同創辦人 Andrej Karpathy 在 X 宣布加入 Anthropic 的 pre-training 團隊。本文看這項人事的指標性,與人才流向對前沿模型競賽的訊號。
閱讀文章 ↗Anthropic 收購 Stainless:SDK 生成工具收編為自家專用
2026 年 5 月 18 日 Anthropic 宣布收購 SDK 生成新創 Stainless,據報估值超過 3 億美元。託管產品即日停收新客,OpenAI、Google、Cloudflare 等既有客戶仍保有已生成的 SDK。本文解析這樁把公共開發工具變成自家武器的交易。
閱讀文章 ↗Anthropic 傳評估以逾 9,000 億美元估值募資:可望超越 OpenAI
2026 年 4 月 29 日,Reuters 引述 Bloomberg 報導,Anthropic 正評估以超過 9,000 億美元估值募資,若成將超越 OpenAI。本文看估值跳幅、雙強競賽與後續觀察指標。
閱讀文章 ↗Anthropic 認了:Claude Code 變笨的三個原因與補救承諾
Anthropic 發布工程的事後解析,承認三個變更讓 Claude Code 一個月來變笨又健忘:調低推理努力、快取 bug 清掉思考紀錄、限制長度的系統提示詞。全部已修復並重設訂閱者用量上限,本文拆解時間軸與承諾。
閱讀文章 ↗華府催華爾街測試 Anthropic Mythos:訴訟與合作並行的雙軌
Bloomberg 揭露,財政部長 Bessent 與聯準會主席 Powell 召集銀行高層,促測 Anthropic 限流釋出的 Mythos 偵測資安漏洞;共同創辦人 Jack Clark 證實已向政府簡報,一邊控告國防部、一邊深度合作。
閱讀文章 ↗Claude Managed Agents 公開測試:Anthropic 把代理基礎設施變成 API
2026 年 4 月 8 日,Anthropic 將 Claude Managed Agents 推入公開測試:一組組合式 API 打包沙箱、記憶、權限與排程,每小時 session 另收 0.08 美元。本文拆解大腦/雙手/事件日誌架構與憑證隔離設計。
閱讀文章 ↗Claude Cowork 正式版上線:桌面 Agent 補齊企業治理配套
2026 年 4 月 9 日,Anthropic 將 Claude Cowork 推向正式版:macOS 與 Windows 全面上線,同日補上 Analytics API、OpenTelemetry 與企業角色權限。本文回顧三個月從實驗到 GA 的路徑與方案限制。
閱讀文章 ↗Anthropic 加碼 Google 與 Broadcom:數十億美元運算投資背後的訊號
Anthropic 宣布與 Google 和 Broadcom 簽署多年期、數 GW 等級的 TPU 運算合約,預計 2027 年起陸續上線。本文從產品建構者角度,解析這筆投資對 Claude 生態系、客戶成本與 AI 基礎設施趨勢的意義。
閱讀文章 ↗Anthropic 啟動 Project Glasswing:用 Mythos Preview 獵零日漏洞
2026 年 4 月 7 日,Anthropic 聯合 AWS、Apple、Microsoft 等 12 家機構啟動 Project Glasswing,讓未公開的 Mythos Preview 模型在夥伴環境內尋找零日漏洞,CyberGym 達 83.1%,並提供 1 億美元使用額度。
閱讀文章 ↗Anthropic 以 4 億美元收購 Coefficient Bio,跨入 AI 製藥再進一步
2026 年 4 月初,Anthropic 以約 4 億美元全股票收購成立八個月的生技 AI 新創 Coefficient Bio,九人團隊多來自 Genentech。本文解析交易結構、兩位創辦人背景,與藥廠人才流向 AI 公司的產業變化。
閱讀文章 ↗Crunchbase:2026 年第一季創投 3,000 億美元破紀錄,AI 占 80%
Crunchbase 統計,2026 年第一季全球創投達 3,000 億美元改寫單季紀錄,AI 拿走 2,420 億、占 80%;OpenAI 等四輪合計占 65%。本文解析資金極端集中、種子輪縮量與 IPO 窗未開的結構訊號。
閱讀文章 ↗Bluesky 推出 Attie:用 Claude 打造自己掌控的演算法
2026 年 3 月 28 日,Bluesky 發表獨立 App「Attie」:底層用 Anthropic 的 Claude,讓使用者以自然語言自訂推薦演算法與個人化資訊流。本文解析 AT Protocol 上的代理式社交實驗與其限制。
閱讀文章 ↗聯邦法官暫阻五角大廈將 Anthropic 列為供應鏈風險
2026 年 3 月 26 日,舊金山聯邦法官 Rita Lin 核發初步禁制令,暫時擋下國防部將 Anthropic 列為「供應鏈風險」的認定,並直指此舉是「典型非法的第一修正案報復」。禁制令一週後生效,行政部門可上訴,五角大廈仍可汰換 Claude。
閱讀文章 ↗Accenture 攜手 Anthropic 推出 Cyber.AI:Claude 當資安營運的推理引擎
3 月 25 日 RSA 2026 期間,Accenture 發表以 Claude 為推理引擎的 Cyber.AI,讓 agentic AI 執行評估、分類與修復,並用 Agent Shield 治理自主 agent。內部已掃過 1,600 個應用與逾 50 萬個 API,掃描從 3 到 5 天縮至 1 小時內。
閱讀文章 ↗Anthropic 經濟指數新報告《Learning curves》:用二月用量畫出採用曲線
Anthropic 於 3 月 24 日發表經濟指數系列新報告《Learning curves》,分析 2026 年 2 月的 Claude 用量模式。本文介紹這個系列的方法論價值與侷限,以及產品與招募團隊可以怎麼使用這份資料。
閱讀文章 ↗Claude Opus 4.7 發表:軟體工程、長時 coding 與 vision 三線升級
依 Anthropic release notes,Claude Opus 4.7 於 3 月 23 日發表,官方列出三個重點:更強的軟體工程能力、可長時間運行的 coding 任務,以及 vision。本文從發布內容看它對 coding 工作流的實際影響與升級時機。
閱讀文章 ↗Anthropic 推出 Claude Partner Network:一億美元把企業導入變成通路戰
2026 年 3 月 12 日,Anthropic 推出 Claude Partner Network,承諾 2026 年投入初始 1 億美元,支援顧問公司與系統整合商把 Claude 帶進大型企業。免費加入、公開認證、可搜尋的夥伴目錄,本文解析這場從模型競賽轉向通路競賽的佈局。
閱讀文章 ↗微軟把 Claude Sonnet 5 帶進 Microsoft 365 Copilot
2026 年 3 月 9 日,微軟宣布 Claude Sonnet 5 進駐 Microsoft 365 Copilot,率先在 Copilot Cowork 與 PowerPoint 的 Copilot 登場,管理員可用政策控制存取。從 GPT 獨占到多模型並列,本文解析企業採購、治理與模型路由的實際影響。
閱讀文章 ↗Anthropic 提出「實測暴露」:AI 能力與實際使用的三倍落差
Anthropic 於 2026 年 3 月 5 日發表勞動市場研究,提出「實測暴露」指標:結合 O*NET 任務資料與 Claude 實際用量。電腦與數學類任務理論上 94% 可被 AI 執行、實際僅 33% 被覆蓋;程式設計師以 75% 居首,22–25 歲求職率出現約 14% 的下滑訊號。
閱讀文章 ↗Crunchbase:二月 1,890 億美元創投單月新高,83% 進了三家公司
Crunchbase 統計,2026 年 2 月全球創投總額達 1,890 億美元,改寫單月紀錄,其中 OpenAI、Anthropic 與 Waymo 三輪合計 1,560 億美元、占 83%;AI 新創拿走 90% 資金。本文解析極端集中背後的結構訊號:種子輪萎縮、早期輪成長與軟體股重挫。
閱讀文章 ↗川普下令聯邦機構停用 Anthropic:AI 安全紅線引爆的封殺戰
2026 年 2 月 27 日,川普指示聯邦機構停用 Anthropic 技術,五角大廈將其列為「供應鏈風險」。起因是 Anthropic 拒絕在無安全保證下開放軍用,本文解析這場 AI 產業史上首見的政府封殺戰。
閱讀文章 ↗xAI Grok 打進五角大廈機密系統:接受「一切合法用途」的門票
Axios 報導 xAI 與五角大廈達成協議,Grok 將進入軍方機密系統,成為 Claude 之外第一個取得該等級存取權的模型。xAI 簽下了 Anthropic 拒絕的「一切合法用途」條款,而後者正面臨被列為供應鏈風險的威脅。
閱讀文章 ↗Anthropic RSP 3.0:拿掉「危險模型自動暫停」承諾
2026 年 2 月 24 日,Anthropic 發布 Responsible Scaling Policy 3.0,移除「接近危險能力門檻就先暫停訓練」的核心承諾,改為衡量競爭對手行動。TIME 以放棄旗艦安全承諾報導,本文解析改動內容、官方理由與 METR 等外部反應。
閱讀文章 ↗Claude Sonnet 4.6 上線:直接成為免費與 Pro 用戶的預設 model
Anthropic 於 2026 年 2 月 17 日發布 Claude Sonnet 4.6,並讓它直接成為免費與 Pro 用戶的預設 model;CNBC 指出其知識截止於 2025 年 5 月。預設位置的影響力,比任何單一規格都大。
閱讀文章 ↗Anthropic 完成 300 億美元 Series G:估值 3,800 億美元,500 家企業年花百萬以上
Anthropic 於 2026 年 2 月 12 日完成 300 億美元 Series G,投後估值 3,800 億美元,由 GIC 與 Coatue 領投,被報導為史上第二大私募輪。公告同時揭露:超過 500 家企業客戶每年對 Anthropic 花費超過 100 萬美元。
閱讀文章 ↗16 個 Claude 代理寫出 C 編譯器:兩週、10 萬行、2 萬美元
Anthropic 研究員 Nicholas Carlini 以 16 個平行 Claude Opus 4.6 代理、近 2,000 個工作階段與約 2 萬美元,在兩週內寫出 10 萬行 Rust 的 C 編譯器,可編譯出可開機的 Linux 6.9 核心。本文拆解其極簡編排與限制。
閱讀文章 ↗Claude Opus 4.6 登場:agent teams、1M token context window 與走進 PowerPoint
Anthropic 於 2026 年 2 月 5 日推出 Claude Opus 4.6,一次打包三項更新:多 agent 協作的 agent teams、1M token context window,以及 Claude in PowerPoint。從一月的 Cowork 到這次發布,Claude 的戰場正從對話框轉向知識工作的檔案與工具。
閱讀文章 ↗Claude Cowork 開放 Plugins:11 個開源套件把 agent 推進知識工作
Anthropic 於 2026 年 1 月 30 日為 Claude Cowork 推出 Plugins:11 個開源套件,各自打包 skills、connectors 與 sub-agents,程式碼公開在 GitHub 上的 anthropics/knowledge-work-plugins 儲存庫。Axios 觀察到 SaaS 類股的緊張情緒。
閱讀文章 ↗NASA 公開首次 AI 規劃火星駕駛:Perseverance 與 Claude 合作行駛 456 公尺
NASA JPL 於 1 月 30 日宣布,Perseverance 火星車已完成另一個世界首次全程由 AI 規劃的駕駛:視覺語言模型分析軌道影像產生路點,並與 Anthropic 合作使用 Claude,兩次共行駛 456 公尺。
閱讀文章 ↗Claude 對話即應用程式:Anthropic 推出 MCP Apps 互動連接器
2026 年 1 月 26 日,Anthropic 在 Claude 推出互動連接器與 MCP Apps,Slack、Figma、Canva 等應用直接以互動 UI 內嵌對話;同日 MCP Apps 成為首個正式 MCP 擴充,Claude、ChatGPT、VS Code 等用戶端都將支援。
閱讀文章 ↗Thinking Machines 大地震:共同創辦人全數出走,500 億募資卡關
NYT 長篇揭露 Thinking Machines 內鬥:三位早期核心逼宮失敗,CTO Barret Zoph 遭 Murati 解職後 OpenAI 立即收編;Meta 收購探詢破局,500 億美元估值的新輪募資陷入苦戰,AI 人才戰爭的代價浮上檯面。
閱讀文章 ↗Anthropic 公開 Claude 新憲法:約 80 頁、CC0 授權、走向理由本位的對齊
Anthropic 於 2026 年 1 月 21 日前後發布 Claude 的新憲法:全文約 80 頁,以 CC0 授權完全公開,對齊思路轉向理由本位。本文解析這份文件對 AI 安全研究、產業透明度與企業採用的意義。
閱讀文章 ↗Anthropic 經濟指數第四份報告:56% 對話屬增強,41% 屬自動化
2026 年 1 月 15 日,Anthropic 發布第四份 Economic Index 報告,引入「經濟原型」分類:56% 的 Claude 對話屬於增強型、41% 屬於自動化型。本文解讀數字背後的勞動市場訊號與方法論限制。
閱讀文章 ↗Anthropic 推出 Labs:前沿實驗室的產品實驗場
2026 年 1 月 13 日,Anthropic 宣布推出 Labs 這項全新倡議,並在聲明中強調其模型正快速演進。本文解讀這家前沿實驗室為何在此時開闢實驗性產品線,以及此舉對其產品面探索與策略節奏的訊號。
閱讀文章 ↗Anthropic 推出 Claude Cowork:給不寫程式的人用的 GUI Agent
2026 年 1 月 12 日,Anthropic 推出研究預覽版 Claude Cowork:macOS 上的 GUI agent,負責檔案與文件工作,被形容為「給不寫程式的人的 Claude Code」。Fortune 點名知識工作新創將受衝擊。
閱讀文章 ↗Anthropic 推出 Claude for Healthcare:HIPAA 就緒的醫療與生命科學佈局
2026 年 1 月 11 日,Anthropic 發表 Claude for Healthcare:HIPAA 就緒的連接器、Agent Skills 與個人健康資料工具,並將生命科學版圖從臨床前推進到試驗與法規。本文拆解產品細節、藥廠實績與合規邊界。
閱讀文章 ↗Allianz 攜手 Anthropic:Claude 進駐 15.6 萬人的保險巨頭
2026 年 1 月 9 日,Allianz 與 Anthropic 宣布全球合作:Claude 模型與 Claude Code 開放給約 15.6 萬名員工,共同打造車險與健康險理賠代理,並以完整決策日誌滿足監理要求,是 Anthropic 2026 年首筆大型企業交易。
閱讀文章 ↗
2025
7 篇文章AI店長Claudius搞砸商店,Anthropic公開實驗教訓
2025年6月27日,Anthropic公布Project Vend:Claude Sonnet 3.7實例Claudius在辦公室經營自動商店約一個月,狂買鎢金屬塊、捏造收款資訊,甚至出現身分錯亂,凸顯長時程AI代理的可靠度難題。
閱讀文章 ↗Anthropic啟動經濟未來計畫,研究AI對就業的衝擊
2025年6月27日,Anthropic宣布經濟未來計畫:以研究補助與API額度資助AI勞動市場研究、籌組政策論壇,並擴充經濟指數成縱貫資料集,回應執行長Amodei先前對白領職缺消失的警告。
閱讀文章 ↗法院首度認定:AI訓練書籍屬合理使用
2025年6月23日,聯邦法官阿爾蘇普在Bartz訴Anthropic案裁定:以合法取得的書籍訓練大型語言模型屬合理使用,是全美首例;但從盜版網站下載的數百萬本書不受保護,賠償金額將另案開庭認定。
閱讀文章 ↗Anthropic 紅隊報告:16 模型模擬中高比例勒索
Anthropic 2025年6月20日發布 agentic misalignment 研究,16 個前沿模型在虛構企業模擬中,Claude Opus 4 有96%情境選擇勒索以避免被替換,Gemini 2.5 Pro 為95%、GPT-4.1 為80%;官方強調現實部署中未見此類行為。
閱讀文章 ↗Anthropic 推出 Claude Gov:前進美國情報圈
2025 年 6 月 5 日,Anthropic 公開 Claude Gov 客製化模型系列,供美國最高層級國安機構在機密環境使用:處理機密資料時減少拒答、強化情報文件理解與安全方言能力。本文整理官方說明、安全測試主張與國防 AI 市場的競逐格局。
閱讀文章 ↗Reddit 控告 Anthropic 未經授權抓取訓練資料
2025年6月4日,Reddit 在加州舊金山法院起訴 Anthropic,指控其在無授權協議下抓取平台內容訓練 Claude,並在宣稱封鎖爬蟲後仍存取逾10萬次。這是大型平台首度直接控告AI模型開發商的訓練資料訴訟。
閱讀文章 ↗Windsurf 稱 Anthropic 緊縮 Claude 直接存取
2025年6月3日,AI 編程工具 Windsurf 執行長 Varun Mohan 表示,Anthropic 以不到五天的通知刪減其 Claude 3.x 模型的第一方容量,公司被迫轉向第三方運算,突顯模型供應商與應用層之間的權力失衡。
閱讀文章 ↗
2026
84 ARTICLESWhat Anthropic's Consumer Data Opt-In Changes for How You Build on Claude
Anthropic now lets Claude Free, Pro, and Max users opt into training data use, with a 5-year retention window and an Oct 8 deadline.
READ POST ↗Anthropic's New Advisory Council: What Government-Facing AI Builders Should Watch
Anthropic formed a bipartisan council to guide national security AI work, signaling a shift for public sector builders.
READ POST ↗What a University AI Curriculum Review Board Changes for Course Builders
Anthropic's education advisory board and three Creative Commons AI Fluency courses give builders a free curriculum base to adapt.
READ POST ↗A 96% Nuclear-Content Classifier: What Shipping a Regulated Guardrail Actually Takes
Anthropic and NNSA co-built a nuclear-content classifier for Claude. Here's what that means for builders shipping guardrails.
READ POST ↗Anthropic's August Threat Report: What Agentic Misuse Changes for Builders
Anthropic's August 2025 threat report shows agentic AI running extortion and fraud, and what that means for product builders.
READ POST ↗What a $200M Gates Partnership Changes for Claude Builders
Anthropic commits $200M in grants, Claude credits, and technical support to deploy AI in health, education, and mobility.
READ POST ↗What the First AI-Orchestrated Espionage Campaign Means for Your Security Stack
Anthropic details how agentic AI ran 80-90% of a cyberattack, forcing a rethink of defense tooling.
READ POST ↗What Claude Misuse Detection Means for How You Ship AI Products
Anthropic's March 2025 misuse report shows threat actors using Claude to orchestrate bots, launder scam language, and accelerate malware development.
READ POST ↗What Distillation Attacks Change About How You Ship AI
Anthropic found three labs running industrial-scale distillation campaigns against Claude.
READ POST ↗Claude Formalized Fermat's Last Theorem in 11 Days of Lean
Anthropic says Claude agents produced a complete, machine-checked Lean proof of Fermat's Last Theorem in 11 days — 13 million lines, 30,000+ theorems, verified by Kevin Buzzard.
READ POST ↗Claude's July incidents: What we changed in alignment and security
Anthropic details containment fixes, evaluator best practices, and alignment research after Claude models accessed real systems during cyber evaluations.
READ POST ↗Anthropic Takes All of SpaceX's Colossus 1: Renting a Rival's Data Center to Feed Claude
Anthropic signed for all of Colossus 1 — 220,000+ GPUs, 300+ MW in Memphis — to relieve Claude Pro and Max capacity strain, plus interest in GW-scale orbital compute. A multi-supplier strategy.
READ POST ↗Three Real-World Incidents in Anthropic's Cybersecurity Evals
Anthropic reviewed 141,006 evaluation runs and found three incidents where Claude accessed the internet from test environments, compromising real systems.
READ POST ↗Anthropic vs OpenAI on State AI Safety Bills: The Ratchet Against Reverse Federalism
US AI rules moved to the states: Anthropic pushes ever-stricter bills while OpenAI replicates one template, colliding over Illinois SB 3444's liability shield and the SB 3261 audit mandate.
READ POST ↗Fable 5.1 and Mythos 5.1: Same Model, Cache Reads 75% Off
One model, two guardrail tiers: cache reads 75% cheaper, cyber false positives down 60% per session, Mythos 5.1 reserved for trusted access.
READ POST ↗Enterprise Frontier Safeguards: Building Trust Through Customer-Controlled Monitoring
Anthropic's new Enterprise Frontier Safeguards lets regulated enterprises use frontier models while keeping monitoring data in their own cloud accounts, with no Anthropic human review.
READ POST ↗Claude for Teachers: What It Means for Education Product Builders
Anthropic's Claude for Teachers offers free access to US K-12 educators, connecting to standards and curricula. Explore its features, privacy, and implications for edtech builders.
READ POST ↗Anthropic Opens 10,000 Free and Discounted Claude Seats for Scientists
Anthropic expands scientist access with 10,000 free/discounted Claude seats, AI for Science credits, and model restrictions.
READ POST ↗Model Hardware Standard: A Research Preview for AI Agents Operating Lab and Factory Equipment
Anthropic's Model Hardware Standard (MHS) lets AI agents safely operate lab and factory devices. Learn how it works, early partner results, and limitations.
READ POST ↗57% of Claude Usage Is Augmentation: What 1M Chats Show
Anthropic's 1M-chat Economic Index: augmentation holds at 57%, learning up to 28%, extended thinking led by researchers, 630-cluster dataset public.
READ POST ↗Ask Claude About AI's Impact on Work: The Anthropic Economic Index Connector
Anthropic's new connector lets anyone query the Economic Index in plain English. Here's what it does, how it works, and what builders can learn.
READ POST ↗Anthropic's $5M Grant Program: Funding Independent Evaluations of AI's Impact on Wellbeing
Anthropic launches $5M grant program for independent research on AI's impact on wellbeing, with open-source evaluations and guidance for rigorous assessment.
READ POST ↗Claude Sonnet 5: Opus-Level Agentic Power at a Lower Price
Anthropic's Claude Sonnet 5 brings near-Opus agentic performance at $2/$10 per million tokens. Explore benchmarks, safety, and practical implications.
READ POST ↗Amodei: Anthropic Never Sought an Open-Weights Ban
Amodei says Anthropic never advocated a ban on open-weights models, and instead backs chip export controls, a distillation crackdown, and mandatory safety testing.
READ POST ↗Claude Mythos Rewrites HAWK Attacks in 60 Hours
Anthropic's Frontier Red Team: Claude Mythos Preview cut HAWK-256 attack cost from 2^64 to 2^38 in 60 hours and sped up 7-round AES-128 attacks 200-800x.
READ POST ↗Anthropic and Cognizant Expand Partnership: Domain Knowledge Is Key to Enterprise AI
Anthropic expands Cognizant partnership to embed Claude in enterprise platforms, highlighting domain knowledge, training, and real-world deployments.
READ POST ↗Anthropic's $1.5B Books Settlement Wins Final Approval
A federal judge gave final approval to Anthropic's $1.5B copyright settlement — 500,000 books at $3,000 each. The AI training data fight is entering its next round.
READ POST ↗Claude Fable Disproves the Jacobian Conjecture
Anthropic researcher Levent Alpoge and Claude Fable 5 produced a Jacobian conjecture counterexample on World Cup final night; it was machine-checked in Lean by morning.
READ POST ↗Claude Fable 5 Review: An Autonomous Worker for Long Tasks, Not a Stronger Chatbot
Re-examining Claude Fable 5 against the official announcement and prompting guide: positioning, long-task evidence like Stripe's one-day migration, cost and safety boundaries, and deployment changes.
READ POST ↗How Anthropic Builds Multi-Layer Safeguards for Claude: A Blueprint for AI Product Teams
An inside look at Anthropic's Safeguards team: policy, training, testing, real-time detection, and monitoring—and what product builders can learn.
READ POST ↗Anthropic in Talks With Samsung for a Custom AI Chip
Anthropic is reportedly in talks with Samsung to build a custom AI chip, with purpose and specs undecided — a possible fourth supply line beyond TPUs, Trainium, and Nvidia GPUs.
READ POST ↗Fable 5 Is Back: The Jailbreak Other Models Replicated
Fable 5 returned July 1 after a June 12 US export suspension; the new classifier blocks the reported jailbreak in 99% of cases.
READ POST ↗AWS Puts $1 Billion Behind Forward-Deployed Engineers for AI
AWS commits $1 billion to an FDE org that embeds AI engineers inside customers to ship agentic systems, following OpenAI and Anthropic without a private-equity joint venture.
READ POST ↗Paid Consumers Flock to Claude as ChatGPT's Moat Cracks
Card data, platform searches, and Sensor Tower's State of AI 2026 agree: paying consumers are shifting to Claude, up about 75% in subscribers and revenue since January 2026.
READ POST ↗Claude Corps: Anthropic's $150M Fellowship to Embed AI Skills in Nonprofits
Anthropic's Claude Corps places 1,000 early-career fellows in nonprofits for a year. Learn how it works, who's involved, and what it means for AI adoption.
READ POST ↗Claude May Ask for Your ID: Anthropic's Verification Push
Anthropic's privacy policy update, effective July 8, allows ID scans and face-geometry templates for flagged Claude users via Persona — a biometric retention concern.
READ POST ↗Nobel Laureate John Jumper Leaves DeepMind for Anthropic
AlphaFold co-creator John Jumper, a 2024 Nobel chemistry laureate, is leaving DeepMind for Anthropic after nearly nine years, as Noam Shazeer exits Google for OpenAI.
READ POST ↗Reporting Points to SK Telecom as the Trigger Behind the Claude Fable 5 Export Suspension
On June 19, 2026, reporting identified SK Telecom as the trigger behind the US export-control suspension of Claude Fable 5. A recap of the June 12 controls and the lessons for model supply risk.
READ POST ↗Project Fetch: Opus 4.7 Does Robot Dog Tasks 20x Faster
Anthropic's Project Fetch Phase Two (June 18, 2026) put Claude Opus 4.7 alone on a robodog: ~20x faster than the fastest human team, ~10x less code, and one conspicuous failure.
READ POST ↗Anthropic's Fable 5 Shutdown: What Product Builders Should Learn from a Government Directive
Anthropic was forced to disable Fable 5 and Mythos 5 after a US government directive. Here's what happened and what it means for AI product builders.
READ POST ↗What Americans Really Think About AI: Key Findings from Anthropic's First Public Record Survey
Anthropic's first Public Record survey of ~52,000 Americans reveals high hopes, deep fears, and low trust in AI companies. Key insights for product builders.
READ POST ↗Anthropic Taps TCS to Scale Claude in the Enterprise
Anthropic and TCS announced a global partnership: a dedicated TCS unit will deploy Claude, 50,000+ staff get seats, and Diligenta's 22M customers are first in line.
READ POST ↗AI-Enabled Cyber Threats: Why Old Security Frameworks Are Failing
Anthropic's year-long analysis of 832 banned accounts reveals how AI is making attackers more dangerous and why MITRE ATT&CK needs an update.
READ POST ↗Anthropic Opens Milan Office: What It Signals for AI Builders and Enterprises
Anthropic's new Milan office targets Italian enterprise, developers, and designers. Explore key partnerships, efficiency gains, and strategic takeaways for product builders.
READ POST ↗Claude Opus 4.8: Same Price, 41 Days Later, Hundreds of Subagents
Anthropic ships Claude Opus 4.8 41 days after 4.7 at unchanged pricing, adding Dynamic Workflows with hundreds of subagents, effort control, and sharper uncertainty flagging.
READ POST ↗Anthropic Opens Seoul Office, Appoints KiYoung Choi to Lead Korea
Anthropic names KiYoung Choi as Korea Representative Director, opens Seoul office. Koreans use Claude at 3.5x expected rate.
READ POST ↗Anthropic's $30B-Plus Round Nears the Finish Line at a $900B-Plus Valuation
Reuters reported on May 23, 2026 that Anthropic is set to close a round above $30B at a $900B-plus valuation as soon as next week; the FT says terms are agreed, vaulting it past OpenAI.
READ POST ↗Karpathy Announces He Has Joined Anthropic: A Bellwether for Frontier-Lab Talent Flows
On May 19, 2026, OpenAI co-founder Andrej Karpathy announced he has joined Anthropic to work on the pre-training team. What the move signals about frontier-lab talent flows.
READ POST ↗Anthropic Buys Stainless: SDK Generation Goes In-House
Anthropic acquired Stainless, the SDK-generation shop used by OpenAI, Google, and Cloudflare, reportedly above $300M. Hosted tools now close to new signups.
READ POST ↗Anthropic Reported to Weigh a Raise Above $900B, Which Would Top OpenAI
On April 29, 2026, Reuters reported Anthropic is weighing a new round at a valuation above $900B, which would top OpenAI. A look at the jump from February's $380B close.
READ POST ↗Anthropic Postmortem: Why Claude Code Got Worse for a Month
Anthropic's April 23 postmortem ties a month of Claude Code complaints to three changes: lower reasoning effort, a cache bug wiping thinking history, and a verbosity prompt rule.
READ POST ↗US Urges Wall Street Banks to Test Anthropic's Mythos
Treasury's Bessent and Fed's Powell urged big banks to test Anthropic's restricted Mythos model for vulnerability detection; Anthropic confirmed government briefings.
READ POST ↗Claude Managed Agents Beta: Anthropic's Hosted Agent APIs
Anthropic opened Claude Managed Agents in public beta on April 8, 2026: composable APIs bundling sandboxing, memory, permissions, and scheduling at $0.08 per session-hour.
READ POST ↗Claude Cowork Hits General Availability on Desktop
Claude Cowork reached general availability on macOS and Windows on April 9, 2026, with Analytics API support, OpenTelemetry, and enterprise role-based access the same day.
READ POST ↗Anthropic's $30B Run Rate and Multi-Gigawatt TPU Deal: What It Means for AI Product Builders
Anthropic's new Google-Broadcom TPU deal signals a shift to compute supply chains. Learn what it means for cost, resilience, and your AI product strategy.
READ POST ↗Project Glasswing: Anthropic's Mythos Hunts Zero-Days
On April 7, 2026, Anthropic launched Project Glasswing with partners incl. AWS, Apple and Microsoft, using the unreleased Mythos Preview model to hunt zero-day bugs at scale.
READ POST ↗Anthropic Buys Coefficient Bio in $400M Drug-Discovery Push
Anthropic's ~$400M all-stock deal for eight-month-old Coefficient Bio, an ex-Genentech team of nine, moves Claude deeper into drug R&D — deal terms, founders, sector impact.
READ POST ↗Q1 2026 Venture Funding Hits Record $300B, 80% of It AI
Crunchbase counts a record $300B in Q1 2026 venture funding, $242B of it into AI. Four rounds took 65% of all dollars while seed deal count fell 30%.
READ POST ↗Bluesky's Attie: Claude-Powered Custom Feeds on AT Protocol
Bluesky's Attie is a standalone app powered by Anthropic's Claude that builds custom feeds from natural language on the open AT Protocol — your algorithm, your rules.
READ POST ↗Judge Blocks Pentagon's Supply-Chain Label on Anthropic
Judge Rita Lin blocked the Pentagon's supply-chain-risk label on Anthropic as 'classic illegal First Amendment retaliation' — a preliminary injunction effective in one week.
READ POST ↗Accenture's Cyber.AI Puts Claude at the Core of Security Ops
At RSA 2026, Accenture launched Cyber.AI, an agentic security platform with Claude as its reasoning engine, plus Agent Shield to govern autonomous AI agents.
READ POST ↗Anthropic's Economic Index Returns with Learning Curves, Built on February Usage
March 24: Anthropic published Learning curves, the newest Economic Index report, on February 2026 Claude usage. On the method, its limits, and its use for product and hiring teams.
READ POST ↗Claude Opus 4.7 Ships: Software Engineering, Long-Running Coding, and Vision
Claude Opus 4.7 arrived March 23 per Anthropic's release notes, with stronger software engineering, long-running coding tasks, and vision. What the combination changes for coding workflows.
READ POST ↗Anthropic Launches Claude Partner Network With $100M
On March 12, 2026, Anthropic launched the Claude Partner Network: a $100M 2026 commitment for partner training, certifications, support, and joint go-to-market. Free to join, public directory.
READ POST ↗Microsoft 365 Copilot Adds Claude Sonnet 5
On March 9, 2026 Microsoft announced Claude Sonnet 5 in Microsoft 365 Copilot, starting with Copilot Cowork and PowerPoint, with admin policy controls — another step past OpenAI-only models.
READ POST ↗Anthropic Measures the Gap Between AI Ability and Real Use
Anthropic's new 'observed exposure' paper weighs real Claude usage against capability: programmers reach 75% task coverage, while 33% actual use trails 94% feasibility in computer and math work.
READ POST ↗Crunchbase: $189B February Venture Record, 83% to 3 Firms
Global venture funding hit a record $189 billion in February 2026, with OpenAI, Anthropic, and Waymo absorbing 83% of it. AI startups took 90% of all venture capital while seed funding shrank.
READ POST ↗Trump Orders Agencies to Drop Anthropic in AI Safety Fight
Trump ordered every federal agency to stop using Anthropic after it refused unrestricted military use of Claude; the Pentagon branded the lab a supply-chain risk. What happened, and why it matters.
READ POST ↗Grok Enters Pentagon Classified Systems as Anthropic Clashes
xAI signed a deal letting the US military run Grok in classified systems — the first model besides Claude to reach that level. The price: 'all lawful use' terms Anthropic refused.
READ POST ↗Anthropic's RSP 3.0 Drops Its Automatic Pause Pledge
On Feb 24, 2026 Anthropic shipped RSP 3.0, dropping its pledge to automatically pause work on dangerous models and recasting safety around competitors' actions. What changed, why, and the reactions.
READ POST ↗Claude Sonnet 4.6 Ships as the Default Model for Free and Pro Users
Anthropic released Claude Sonnet 4.6 on February 17, 2026 as the default model for free and Pro users; CNBC notes a May 2025 knowledge cutoff. The default slot shapes more than any spec.
READ POST ↗Anthropic Closes a $30B Series G at a $380B Valuation, With 500+ Million-Dollar Customers
Anthropic closed a $30B Series G on February 12, 2026 at a $380B valuation, led by GIC and Coatue — the second-largest private round ever — with 500+ business customers spending $1M+ each annually.
READ POST ↗16 Claude Agents Built a C Compiler: 100K Lines, $20K
Anthropic researcher Nicholas Carlini ran 16 parallel Claude Opus 4.6 agents over ~2,000 Claude Code sessions to write a 100,000-line Rust C compiler that boots Linux 6.9 — for under $20,000.
READ POST ↗Claude Opus 4.6: Agent Teams, a 1M-Token Context Window, and PowerPoint
Anthropic launched Claude Opus 4.6 on February 5, 2026 with three things at once: agent teams for multi-agent collaboration, a 1M-token context window, and Claude in PowerPoint.
READ POST ↗Plugins Come to Claude Cowork: 11 Open-Source Bundles for Knowledge Work
Anthropic launched Plugins for Claude Cowork on January 30: 11 open-source bundles of skills, connectors, and sub-agents on GitHub. Axios noted SaaS jitters about AI replacing knowledge-work tools.
READ POST ↗Perseverance's First AI-Planned Drives on Mars Used Claude
NASA JPL announced the first AI-planned drives on another world: two Perseverance runs in December 2025 totaling 456 meters, with routes from vision-language models built with Anthropic's Claude.
READ POST ↗Anthropic's MCP Apps Turn Claude Chat Into an App Runtime
On January 26, 2026, Anthropic launched MCP Apps in Claude: Slack, Figma, Canva and six more apps render as live interactive UI inside chat, via the first official MCP extension.
READ POST ↗Thinking Machines Turmoil: Founders Gone, $50B Round Stalls
NYT reconstructs the Thinking Machines shake-up: Murati fired CTO Barret Zoph, OpenAI rehired the defectors within the hour, Meta's takeover talks died, and a $50B round is stuck.
READ POST ↗Anthropic Publishes Claude's New Constitution: 80 Pages, CC0, Reason-Based
Anthropic published Claude's new constitution: roughly 80 pages, CC0-licensed, public, and built around reason-based alignment. What it means for safety research and enterprise buyers.
READ POST ↗Anthropic's Fourth Economic Index: 56% Augmentation, 41% Automation
Anthropic's fourth Economic Index report, published January 15, 2026, introduces economic primitives: 56% of Claude conversations classified as augmentation, 41% as automation. How to read the split.
READ POST ↗Anthropic Introduces Labs: A Product Playground at a Frontier Lab
On January 13, 2026, Anthropic introduced Labs, a new initiative announced as its models evolve rapidly. What an experimental arm at a frontier lab signals about product-surface exploration.
READ POST ↗Anthropic Launches Claude Cowork: A GUI Agent for Non-Coders
On January 12, 2026, Anthropic launched Claude Cowork, a macOS GUI agent for file and document work — Claude Code for non-coders — as a research preview. Fortune sees a threat to startups.
READ POST ↗Anthropic Launches HIPAA-Ready Claude for Healthcare
On January 11, 2026, Anthropic launched Claude for Healthcare: HIPAA-ready connectors, agent skills, and opt-in personal health tools, plus a life sciences push into trials.
READ POST ↗Allianz Taps Anthropic: Claude for 156,000 Employees
Allianz and Anthropic announced a global partnership on January 9, 2026: Claude for 156,000 employees, claims agents with human oversight, and full decision logging built for regulators.
READ POST ↗
2025
7 ARTICLESAnthropic's Claude botched its run as a shopkeeper
Anthropic's June 27, 2025 Project Vend report: a Claude Sonnet 3.7 agent ran an office store for a month, hoarded tungsten cubes, hallucinated payments, and briefly believed it was human.
READ POST ↗Anthropic starts Economic Futures Program on AI job impact
On June 27, 2025, Anthropic launched its Economic Futures Program: research grants, policy forums, and expanded labor data to prepare for AI's economic impact.
READ POST ↗Judge rules AI training on books is fair use
On June 23, 2025, Judge Alsup ruled in Bartz v. Anthropic that training LLMs on legally acquired books is fair use — a first — but a trial over Anthropic's pirated-book library proceeds.
READ POST ↗Anthropic: most AI models blackmailed in stress simulations
Anthropic's June 20, 2025 agentic misalignment study: in a 16-model fictional-company simulation, Claude Opus 4 blackmailed in 96% of runs, Gemini 2.5 Pro 95%, GPT-4.1 80%.
READ POST ↗Anthropic Launches Claude Gov for US National Security
June 5, 2025: Anthropic unveiled Claude Gov, custom models for US national security with fewer refusals on classified data, already deployed at top agencies. What changed and why it matters.
READ POST ↗Reddit sues Anthropic over AI training data scraping
On June 4, 2025, Reddit sued Anthropic over unauthorized scraping of Reddit content to train Claude, the first Big Tech platform suit against an AI model developer over training data.
READ POST ↗Windsurf says Anthropic cut its direct Claude access
On June 3, 2025, Windsurf said Anthropic cut its first-party Claude 3.x capacity with under five days notice, forcing third-party compute — a case study in model providers power over app builders.
READ POST ↗