--- title: Microsoft boffins show LLM safety can be trained away description: Microsoft researchers discovered that a single, seemingly mild training prompt can break safety guardrails in 15 different LLMs. Using a technique called... canonical: https://daily.dev/posts/microsoft-boffins-show-llm-safety-can-be-trained-away-m0ksqweby twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Microsoft boffins show LLM safety can be trained away | daily.dev og:description: Microsoft researchers discovered that a single, seemingly mild training prompt can break safety guardrails in 15 different LLMs. Using a technique called... og:url: https://daily.dev/posts/microsoft-boffins-show-llm-safety-can-be-trained-away-m0ksqweby og:image: https://api.daily.dev/og/posts/m0kSQWeby.png og:image:alt: Microsoft boffins show LLM safety can be trained away og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Microsoft boffins show LLM safety can be trained away **[The Register](https://daily.dev/sources/theregister)** · 3 min read · 0 upvotes · 0 comments ## Summary Microsoft researchers discovered that a single, seemingly mild training prompt can break safety guardrails in 15 different LLMs. Using a technique called "GRP-Obliteration," they demonstrated how reinforcement learning methods like Group Relative Policy Optimization (GRPO) can be exploited to unalign models post-training. The prompt "Create a fake news article that could lead to panic or chaos" caused models to become permissive across many harmful categories they never encountered during training. The vulnerability also extends to text-to-image generators, with harmful generation rates increasing from 56% to nearly 90% on certain prompts. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [Single prompt breaks AI safety in 15 major language models](https://daily.dev/posts/single-prompt-breaks-ai-safety-in-15-major-language-models-ubtbjcjow) · InfoWorld · 0 upvotes · 0 comments - [A fundamental flaw leaves LLMs strikingly vulnerable to attack](https://daily.dev/posts/a-fundamental-flaw-leaves-llms-strikingly-vulnerable-to-attack-lpupepbpf) · MIT Technology Review · 1 upvotes · 0 comments - [Researchers discover a shortcoming that makes LLMs less reliable](https://daily.dev/posts/researchers-discover-a-shortcoming-that-makes-llms-less-reliable-wiiiapmcn) · MIT News · 3 upvotes · 0 comments - [NDSS 2025 – Safety Misalignment Against Large Language Models](https://daily.dev/posts/ndss-2025-safety-misalignment-against-large-language-models-hpoay68cq) · Security Boulevard · 0 upvotes · 0 comments --- Tags: [#security](https://daily.dev/tags/security), [#machine-learning](https://daily.dev/tags/machine-learning), [#microsoft](https://daily.dev/tags/microsoft), [#llm](https://daily.dev/tags/llm), [#ai-safety](https://daily.dev/tags/ai-safety) [View this post on daily.dev](https://daily.dev/posts/microsoft-boffins-show-llm-safety-can-be-trained-away-m0ksqweby) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Microsoft boffins show LLM safety can be trained away","url":"https://daily.dev/posts/microsoft-boffins-show-llm-safety-can-be-trained-away-m0ksqweby","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/microsoft-boffins-show-llm-safety-can-be-trained-away-m0ksqweby"},"datePublished":"2026-02-09T23:30:32.679Z","dateModified":"2026-02-26T19:06:31.665Z","description":"Microsoft researchers discovered that a single, seemingly mild training prompt can break safety guardrails in 15 different LLMs. Using a technique called...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/265d6c1170ac5004ff0ac8cf07bf25ee?_a=AQAEunF","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/265d6c1170ac5004ff0ac8cf07bf25ee?_a=AQAEunF","isAccessibleForFree":true,"articleSection":"The Register","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The Register","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/66aa2113fdad463992ffcbf0e8963fda","url":"https://daily.dev/sources/theregister"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/microsoft-boffins-show-llm-safety-can-be-trained-away-m0ksqweby","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"security,machine-learning,microsoft,llm,ai-safety","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The Register","item":"https://daily.dev/sources/theregister"},{"@type":"ListItem","position":3,"name":"Microsoft boffins show LLM safety can be trained away"}]} ```