{"id":633413,"date":"2026-06-11T17:09:15","date_gmt":"2026-06-11T17:09:15","guid":{"rendered":"https:\/\/www.newsbeep.com\/uk\/633413\/"},"modified":"2026-06-11T17:09:15","modified_gmt":"2026-06-11T17:09:15","slug":"anthropic-apologizes-for-invisible-claude-fable-guardrails","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/uk\/633413\/","title":{"rendered":"Anthropic apologizes for invisible Claude Fable guardrails"},"content":{"rendered":"<p class=\"duet--article--dangerously-set-cms-markup duet--article--standard-paragraph _1044qizi _18mzr4b1 _18mzr4b0 _19wv7tc1\">Anthropic has apologized for stealthily throttling its <a href=\"https:\/\/www.theverge.com\/news\/946725\/anthropic-releases-claude-fable-5-mythos\" rel=\"nofollow noopener\" target=\"_blank\">new AI model, Claude Fable 5<\/a>, with hidden guardrails that undermine both researchers and rivals using it to develop competing systems. The company says it is reversing course and will be more transparent about when the restrictions kick in, even if that means Fable refuses more queries.<\/p>\n<p class=\"duet--article--dangerously-set-cms-markup duet--article--standard-paragraph _1044qizi _18mzr4b1 _18mzr4b0 _19wv7tc1\">Fable is the first widely available model in Anthropic\u2019s Mythos class of AI systems, a group the company has spent months warning are <a href=\"https:\/\/www.theverge.com\/ai-artificial-intelligence\/917644\/anthropic-claude-mythos-breach-humiliation\" rel=\"nofollow noopener\" target=\"_blank\">too dangerous for public release<\/a>. Anthropic says it has addressed some of those risks by launching Fable with safeguards that prevent it from responding to certain \u201chigh-risk\u201d queries. One of the areas Anthropic <a href=\"https:\/\/www.anthropic.com\/news\/claude-fable-5-mythos-5\" rel=\"nofollow noopener\" target=\"_blank\">said<\/a> it would restrict Fable\u2019s responses is distillation, a technique for training smaller AI models using the outputs of larger ones.<\/p>\n<p class=\"duet--article--dangerously-set-cms-markup duet--article--standard-paragraph _1044qizi _18mzr4b1 _18mzr4b0 _19wv7tc1\">In <a href=\"https:\/\/www-cdn.anthropic.com\/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf\" rel=\"nofollow noopener\" target=\"_blank\">Fable\u2019s system card<\/a> \u2014 a public document AI developers release to explain how a system works \u2014 Anthropic said it would handle queries it believed were distillation attempts by altering and degrading the model\u2019s answers directly. Users would not be notified that they had triggered the safety measure or informed that the responses had been changed.<\/p>\n<p class=\"duet--article--dangerously-set-cms-markup duet--article--standard-paragraph _1044qizi _18mzr4b1 _18mzr4b0 _19wv7tc1\">Anthropic <a href=\"https:\/\/x.com\/ClaudeDevs\/status\/2064949876463645026?s=20\" rel=\"nofollow\">said<\/a> it is now changing its approach to distillation: Queries will now fall back to Claude Opus 4.8, Anthropic\u2019s <a href=\"https:\/\/www.theverge.com\/ai-artificial-intelligence\/939094\/anthropic-claude-4-8-opus-honesty-effort\" rel=\"nofollow noopener\" target=\"_blank\">previous flagship model<\/a>, the company said in a post on X. Anthropic will prominently tell users too: \u201cYou will see this every time it happens.\u201d<\/p>\n<p class=\"duet--article--dangerously-set-cms-markup duet--article--standard-paragraph _1044qizi _18mzr4b1 _18mzr4b0 _19wv7tc1\">This is similar to how Fable handles queries in other high-risk areas. When safety features are triggered in areas like biology, chemistry, and cybersecurity, queries are routed through Opus 4.8 unless they are blocked outright under the company\u2019s broader safety rules, such as those covering drugs, weapons, or other prohibited content. In some cases, notably biology, the safeguards have been calibrated so broadly that Fable is <a href=\"https:\/\/www.theverge.com\/ai-artificial-intelligence\/947973\/fable-wont-answer-basic-biology-questions\" rel=\"nofollow noopener\" target=\"_blank\">practically unusable for even basic queries<\/a>, something Anthropic acknowledged in a comment to The Verge.<\/p>\n<p class=\"duet--article--dangerously-set-cms-markup duet--article--standard-paragraph _1044qizi _18mzr4b1 _18mzr4b0 _19wv7tc1\">\u201cVisible safeguards can be probed, so they have to be robust, which takes time to get right,\u201d Anthropic wrote. \u201cInvisible safeguards can be targeted more narrowly, allowing us to ship quickly with very few false positives. We went with invisible safeguards for this reason\u2014and that was the wrong tradeoff. You should have visibility into the safeguards we have in place, and why. We\u2019re sorry for not getting the balance right.\u201d<\/p>\n<p class=\"duet--article--dangerously-set-cms-markup duet--article--standard-paragraph _1044qizi _18mzr4b1 _18mzr4b0 _19wv7tc1\">The change follows <a href=\"https:\/\/www.wired.com\/story\/anthropic-responds-to-backlash-on-claudes-secret-sabotage-on-ai-research\/\" rel=\"nofollow noopener\" target=\"_blank\">intense backlash from the AI research community<\/a> over Anthropic\u2019s decision to silently limit users suspected of trying to distill Fable into competing models \u2014 a safeguard critics warned could also affect third parties trying to evaluate the frontier model. In the system card, Anthropic said newer models\u2019 ability to accelerate AI development justified targeting those requests, noting that \u201cusing Claude to develop competing models already violates our Terms of Service.\u201d Anthropic has previously <a href=\"https:\/\/www.theverge.com\/ai-artificial-intelligence\/883243\/anthropic-claude-deepseek-china-ai-distillation\" rel=\"nofollow noopener\" target=\"_blank\">accused<\/a> Chinese rivals like DeepSeek of unfairly distilling its models on an \u201cindustrial\u201d scale.<\/p>\n","protected":false},"excerpt":{"rendered":"Anthropic has apologized for stealthily throttling its new AI model, Claude Fable 5, with hidden guardrails that undermine&hellip;\n","protected":false},"author":2,"featured_media":633414,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5],"tags":[554,1119,84,59,50,56,54,55],"class_list":["post-633413","post","type-post","status-publish","format-standard","has-post-thumbnail","category-business","tag-ai","tag-anthropic","tag-business","tag-gb","tag-news","tag-uk","tag-united-kingdom","tag-unitedkingdom"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/633413","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/comments?post=633413"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/633413\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media\/633414"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media?parent=633413"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/categories?post=633413"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/tags?post=633413"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}