[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-anthropics-model-shutdown-safety-can-bite-back-en":3,"article-related-anthropics-model-shutdown-safety-can-bite-back-en":33,"series-industry-d2810cd9-a360-4466-a3a3-5a953daea1b1":80},{"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":25,"views":29,"created_at":30,"published_at":31,"topic_cluster_id":32},"d2810cd9-a360-4466-a3a3-5a953daea1b1","anthropics-model-shutdown-safety-can-bite-back-en","Anthropic’s model shutdown shows safety can bite back","\u003Cp data-speakable=\"summary\">\u003Ca href=\"\u002Ftag\u002Fanthropic\">Anthropic\u003C\u002Fa>’s strongest AI models were shut down worldwide after a U.S. government order.\u003C\u002Fp>\u003Cp>Anthropic’s latest clash with Washington shows what happens when safety claims meet export control power. The company says it received the order at 5:21 p.m. ET and had to disable two models for every user worldwide, even though the directive was aimed at foreign nationals.\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>Item\u003C\u002Fth>\u003Cth>Status\u003C\u002Fth>\u003Cth>Why it mattered\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>Claude Mythos 5\u003C\u002Ftd>\u003Ctd>Disabled worldwide\u003C\u002Ftd>\u003Ctd>Most capable model, tightly restricted before the order\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Claude Fable 5\u003C\u002Ftd>\u003Ctd>Disabled worldwide\u003C\u002Ftd>\u003Ctd>General-release model with guardrails\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Project Glasswing\u003C\u002Ftd>\u003Ctd>Ended access\u003C\u002Ftd>\u003Ctd>Controlled program for vetted security users\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>1. Claude Mythos 5\u003C\u002Fh2>\u003Cp>Mythos was Anthropic’s most capable model and the one the company had treated as too risky for broad release. The company said it had kept it under tight control because of its ability to find software vulnerabilities across major platforms.\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781932667519-4tfs.png\" alt=\"Anthropic’s model shutdown shows safety can bite back\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>That caution made Mythos useful for defensive security work, but it also made it the centerpiece of the shutdown. Anthropic says the U.S. government’s order now forces it to turn the model off for all users worldwide, not just the foreign nationals named in the export control action.\u003C\u002Fp>\u003Cul>\u003Cli>Previewed in early April\u003C\u002Fli>\u003Cli>Restricted to a small set of vetted users\u003C\u002Fli>\u003Cli>Used to test security flaws in operating systems and browsers\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>2. Claude Fable 5\u003C\u002Fh2>\u003Cp>Fable 5 was Anthropic’s attempt to package Mythos-level capability for public use. The company said it added guardrails that block high-risk areas such as \u003Ca href=\"\u002Ftag\u002Fcybersecurity\">cybersecurity\u003C\u002Fa> and biology, making the model safe enough for general release.\u003C\u002Fp>\u003Cp>That release lasted only three days before the shutdown order landed. Anthropic says the government’s concern appears to be a claimed jailbreak, while the company argues the issue is narrow and not unique among frontier models.\u003C\u002Fp>\u003Cul>\u003Cli>Released three days before the shutdown\u003C\u002Fli>\u003Cli>Marketed as the public version of Mythos\u003C\u002Fli>\u003Cli>Anthropic says the model still had independent safety classifiers\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>3. Project Glasswing\u003C\u002Fh2>\u003Cp>Project Glasswing was Anthropic’s answer to the problem of giving powerful models to security teams without opening them up to everyone. The program shared Mythos with about 50 vetted organizations for defensive cybersecurity work.\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781932660207-myoq.png\" alt=\"Anthropic’s model shutdown shows safety can bite back\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>Those organizations included \u003Ca href=\"https:\u002F\u002Fwww.amazon.com\u002F\" target=\"_blank\" rel=\"noopener noreferrer\">Amazon\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fwww.apple.com\u002F\" target=\"_blank\" rel=\"noopener noreferrer\">Apple\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fwww.google.com\u002F\" target=\"_blank\" rel=\"noopener noreferrer\">Google\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fwww.microsoft.com\u002F\" target=\"_blank\" rel=\"noopener noreferrer\">Microsoft\u003C\u002Fa>, and \u003Ca href=\"https:\u002F\u002Fwww.crowdstrike.com\u002F\" target=\"_blank\" rel=\"noopener noreferrer\">CrowdStrike\u003C\u002Fa>. Anthropic’s shutdown means that controlled access path is now gone too, at least for the affected models.\u003C\u002Fp>\u003Cul>\u003Cli>About 50 vetted organizations took part\u003C\u002Fli>\u003Cli>Focused on defensive cybersecurity\u003C\u002Fli>\u003Cli>Built to limit exposure while preserving utility\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>4. The U.S. government export control order\u003C\u002Fh2>\u003Cp>The government framed the move as an export control action, but Anthropic says the effect was much broader than the stated target. The company says the order arrived Friday evening and required a global cutoff, which it complied with even while disputing the reasoning.\u003C\u002Fp>\u003Cp>Anthropic says the underlying concern was a possible jailbreak of Fable 5, described as prompting the model to inspect a codebase and identify flaws. The company argues that capability is already common in other public models, including \u003Ca href=\"https:\u002F\u002Fopenai.com\u002F\" target=\"_blank\" rel=\"noopener noreferrer\">OpenAI\u003C\u002Fa>’s GPT-5.5.\u003C\u002Fp>\u003Cul>\u003Cli>Received Friday at 5:21 p.m. ET\u003C\u002Fli>\u003Cli>Applied to all users worldwide\u003C\u002Fli>\u003Cli>Left Anthropic’s other models untouched\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>5. The business risk for frontier AI labs\u003C\u002Fh2>\u003Cp>Anthropic’s complaint is not just about this one order. The company warned that if a narrow jailbreak is enough to force a recall, then frontier model launches could become far harder for every major lab.\u003C\u002Fp>\u003Cp>That matters because Anthropic has built much of its brand on caution, and it is widely expected to pursue an \u003Ca href=\"\u002Ftag\u002Fipo\">IPO\u003C\u002Fa> this year. The episode may now turn its safety-first posture into a business risk, especially if regulators decide that public warnings are proof the model is too dangerous to leave in circulation.\u003C\u002Fp>\u003Cul>\u003Cli>Could slow future model releases\u003C\u002Fli>\u003Cli>Raises questions about how safety claims affect regulation\u003C\u002Fli>\u003Cli>Highlights the tension between product launches and oversight\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>How to decide\u003C\u002Fh2>\u003Cp>If you want the core event, focus on the government order and the shutdown of Fable 5 and Mythos 5. If you want the bigger lesson, watch how Anthropic’s own warnings about danger may have helped trigger the scrutiny it feared.\u003C\u002Fp>\u003Cp>For readers tracking AI policy, this is a case study in how safety messaging, export controls, and commercial pressure can collide. For product teams, it is a warning that saying a model is unusually risky can change how regulators treat it later.\u003C\u002Fp>","Anthropic’s safest models were shut down worldwide after a U.S. government order, exposing the cost of warning too loudly.","techcrunch.com","https:\u002F\u002Ftechcrunch.com\u002F2026\u002F06\u002F12\u002Fanthropics-safety-warnings-may-have-just-backfired-the-government-has-pulled-the-plug-on-its-most-powerful-ai\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781932667519-4tfs.png","industry","en","3f6c0c0f-4d9f-4222-96df-307ea939e4cd",[17,18,19,20,21,22,23,24],"Anthropic","Claude Fable 5","Claude Mythos 5","AI safety","export controls","U.S. government","frontier models","cybersecurity",[26,27,28],"Anthropic says it had to disable Claude Mythos 5 and Claude Fable 5 worldwide after a U.S. government order.","The company argues the government acted over a narrow jailbreak claim that is not unique to its models.","The case shows how aggressive safety warnings can invite the same scrutiny they are meant to prevent.",0,"2026-06-20T05:17:22.132183+00:00","2026-06-20T05:17:22.124+00:00","520d4d46-4f94-4abd-a36f-af3a6c940a98",{"tags":34,"relatedLang":39,"relatedPosts":43},[35,37],{"name":17,"slug":36},"anthropic",{"name":20,"slug":38},"ai-safety",{"id":15,"slug":40,"title":41,"language":42},"anthropics-model-shutdown-safety-can-bite-back-zh","Anthropic 模型下架：安全警告反噬了自己","zh",[44,50,56,62,68,74],{"id":45,"slug":46,"title":47,"cover_image":48,"image_url":48,"created_at":49,"category":13},"2c317df8-4070-4c74-bab5-48f79fe2860e","claude-vs-gpt-vs-gemini-coding-benchmark-leaderboard-en","Claude vs GPT vs Gemini: Coding Benchmark Leaderboard","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781939876788-ivgw.png","2026-06-20T07:17:35.473285+00:00",{"id":51,"slug":52,"title":53,"cover_image":54,"image_url":54,"created_at":55,"category":13},"e018e62b-a712-4e2c-aee6-21fb492b993a","clip-converter-rivals-faster-safer-2026-en","Clip Converter’s 2026 rivals are faster and safer","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781935360953-hmtl.png","2026-06-20T06:02:19.249403+00:00",{"id":57,"slug":58,"title":59,"cover_image":60,"image_url":60,"created_at":61,"category":13},"e5877eb6-413d-46f3-b91e-3c4139b5e1f9","openai-sora-shutdown-unit-economics-en","OpenAI’s Sora shutdown proves hype can’t outrun unit economics","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781933563230-w87h.png","2026-06-20T05:32:17.714689+00:00",{"id":63,"slug":64,"title":65,"cover_image":66,"image_url":66,"created_at":67,"category":13},"58f4c299-67b3-4f19-bc1d-7bb6ef0db0f2","boy-george-ai-vs-taylor-swift-rerecordings-en","Boy George AI vs Taylor Swift rerecordings","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781931778008-edt4.png","2026-06-20T05:02:33.338344+00:00",{"id":69,"slug":70,"title":71,"cover_image":72,"image_url":72,"created_at":73,"category":13},"55699da8-8f47-4348-81c3-65cd969debd3","four-music-datasets-ai-music-training-en","Four music datasets are shaping AI music training","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781930863975-psfj.png","2026-06-20T04:47:22.638469+00:00",{"id":75,"slug":76,"title":77,"cover_image":78,"image_url":78,"created_at":79,"category":13},"27db8279-3710-431a-a86b-20ca47af3a15","deezer-ai-music-detector-playlists-transparency-en","Deezer is right to expose AI music in playlists","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781929965653-f5r4.png","2026-06-20T04:32:18.064695+00:00",[81,86,91,96,101,106,111,116,121,126],{"id":82,"slug":83,"title":84,"created_at":85},"d35a1bd9-e709-412e-a2df-392df1dc572a","ai-impact-2026-developments-market-en","AI's Impact in 2026: Key Developments and Market Shifts","2026-03-25T16:20:33.205823+00:00",{"id":87,"slug":88,"title":89,"created_at":90},"5ed27921-5fd6-492e-8c59-78393bf37710","trumps-ai-legislative-framework-en","Trump's AI Legislative Framework: What's Inside?","2026-03-25T16:22:20.005325+00:00",{"id":92,"slug":93,"title":94,"created_at":95},"e454a642-f03c-4794-b185-5f651aebbaca","nvidia-gtc-2026-key-highlights-innovations-en","NVIDIA GTC 2026: Key Highlights and Innovations","2026-03-25T16:22:47.882615+00:00",{"id":97,"slug":98,"title":99,"created_at":100},"0ebb5b16-774a-4922-945d-5f2ce1df5a6d","claude-usage-diversifies-learning-curves-en","Claude Usage Diversifies, Learning Curves Emerge","2026-03-25T16:25:50.770376+00:00",{"id":102,"slug":103,"title":104,"created_at":105},"69934e86-2fc5-4280-8223-7b917a48ace8","openclaw-ai-commoditization-concerns-en","OpenClaw's Rise Raises Concerns of AI Model Commoditization","2026-03-25T16:26:30.582047+00:00",{"id":107,"slug":108,"title":109,"created_at":110},"b4b2575b-2ac8-46b2-b90e-ab1d7c060797","google-gemini-ai-rollout-2026-en","Google's Gemini AI Rollout Extended to 2026","2026-03-25T16:28:14.808842+00:00",{"id":112,"slug":113,"title":114,"created_at":115},"6e18bc65-42ae-4ad0-b564-67d7f66b979e","meta-llama4-fabricated-results-scandal-en","Meta's Llama 4 Scandal: Fabricated AI Test Results Unveiled","2026-03-25T16:29:15.482836+00:00",{"id":117,"slug":118,"title":119,"created_at":120},"bf888e9d-08be-4f47-996c-7b24b5ab3500","accenture-mistral-ai-deployment-en","Accenture and Mistral AI Team Up for AI Deployment","2026-03-25T16:31:01.894655+00:00",{"id":122,"slug":123,"title":124,"created_at":125},"5382b536-fad2-49c6-ac85-9eb2bae49f35","mistral-ai-high-stakes-2026-en","Mistral AI: Facing High Stakes in 2026","2026-03-25T16:31:39.941974+00:00",{"id":127,"slug":128,"title":129,"created_at":130},"9da3d2d6-b669-4971-ba1d-17fdb3548ed5","cursors-meteoric-rise-pressures-en","Cursor's Meteoric Rise Faces Industry Pressures","2026-03-25T16:32:21.899217+00:00"]