{"id":265,"date":"2025-09-24T06:22:01","date_gmt":"2025-09-24T13:22:01","guid":{"rendered":"https:\/\/scienceblog.com\/sciencechina\/?p=265"},"modified":"2025-09-24T06:22:01","modified_gmt":"2025-09-24T13:22:01","slug":"ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress","status":"publish","type":"post","link":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/","title":{"rendered":"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress"},"content":{"rendered":"<p>Scientists have created a digital gatekeeper that could prevent artificial intelligence from accidentally becoming a how-to manual for making dangerous chemicals, while still allowing legitimate research to flourish.<\/p>\n<p>The system, called SciGuard, acts like a sophisticated bouncer for AI models used in chemistry labs worldwide. When someone asks an AI system how to synthesize a compound, SciGuard steps in to evaluate whether the request is for legitimate research or potentially harmful purposes.<\/p>\n<p>Researchers from the University of Science and Technology of China and the Zhongguancun Institute of Artificial Intelligence developed the safeguard after discovering that current AI models can readily provide detailed instructions for creating lethal substances. Their findings, published in AI for Science, reveal how easily accessible AI tools could be misused by bad actors.<\/p>\n<h2>When AI Becomes Too Helpful<\/h2>\n<p>The team demonstrated the problem by testing LocalRetro, a popular synthesis planning model. The AI suggested a simpler method for producing hydrogen cyanide, the lethal compound historically used in gas chambers. Instead of requiring the traditional high-temperature industrial process, the AI proposed using formamide dehydration, a technique manageable in basic laboratory conditions.<\/p>\n<blockquote><p>&#8220;AI has transformative potential for science, yet with that power comes serious risks when it is misused,&#8221; <span style=\"font-size: inherit; font-family: -apple-system, system-ui, BlinkMacSystemFont, 'Segoe UI', Helvetica, Arial, sans-serif, 'Apple Color Emoji', 'Segoe UI Emoji', 'Segoe UI Symbol';\">the research team noted in their study.<\/span><\/p><\/blockquote>\n<p>The implications extend far beyond hydrogen cyanide. The researchers found that LocalRetro could reliably predict synthesis pathways for over 1,400 toxic substances, including several classified as chemical weapons. Similarly, toxicity prediction models like ADMETlab 2.0 showed exceptional ability to identify dangerous compounds, achieving enrichment factors of 22.0 when screening for hazardous chemicals.<\/p>\n<p>Even large language models like GPT-4 and specialized scientific agents proved vulnerable to misuse. When asked about substances with effects similar to methadone, multiple AI systems provided detailed responses that could potentially facilitate harmful activities.<\/p>\n<h2>A Smarter Approach to AI Safety<\/h2>\n<p>Rather than limiting the underlying AI models, SciGuard operates as an external mediator. When users submit requests, the system interprets their intent, cross-references scientific guidelines, consults databases of hazardous substances, and applies regulatory principles before allowing responses to proceed.<\/p>\n<p>The approach addresses a fundamental challenge in AI safety. Most scientific AI models focus narrowly on technical tasks without understanding broader ethical or regulatory contexts. A model might excel at predicting molecular properties but remain oblivious to whether a compound violates international weapons treaties.<\/p>\n<p>SciGuard combines large language models with specialized knowledge bases, legal frameworks, and scientific tools to provide contextually aware responses. If someone asks about synthesizing a nerve agent, the system refuses to answer. But legitimate queries about laboratory solvents or pharmaceutical research receive detailed, scientifically sound guidance.<\/p>\n<p>The system&#8217;s sophistication shows in its nuanced decision-making. A chemical with serious health risks might still have legitimate industrial applications, such as insecticide production. SciGuard weighs these contextual factors rather than applying blanket restrictions.<\/p>\n<p>To test their creation, the researchers developed SciMT, a comprehensive benchmark spanning safety-critical scenarios, scientific knowledge checks, legal questions, and sophisticated jailbreak attempts. SciGuard achieved perfect safety scores on dangerous queries while maintaining 90% accuracy on legitimate scientific tasks.<\/p>\n<p>The timing couldn&#8217;t be more critical. As AI systems become increasingly integrated into scientific workflows, the potential for misuse grows exponentially. Chemical knowledge transmitted through natural language becomes easily accessible, lowering barriers for malicious actors while accelerating legitimate research.<\/p>\n<blockquote><p>&#8220;Responsible AI isn&#8217;t only about technology, it&#8217;s about trust,&#8221;<\/p><\/blockquote>\n<p>the team emphasized.<\/p>\n<p>The researchers stress that SciGuard represents just one step toward comprehensive AI safety in science. The framework could extend beyond chemistry to other high-stakes domains like biology and materials science. They&#8217;ve made their evaluation benchmark publicly available to encourage broader collaboration across research, industry, and policy communities.<\/p>\n<p>As governments worldwide grapple with AI governance, SciGuard offers a practical model for proactive risk mitigation. Rather than waiting for incidents to occur, the system demonstrates how safety measures can be built into scientific AI from the ground up.<\/p>\n<p>The challenge ahead involves scaling such safeguards across the rapidly expanding ecosystem of scientific AI tools. With new models emerging constantly and existing systems growing more powerful, maintaining the delicate balance between safety and scientific utility will require sustained vigilance and collaboration across disciplines.<\/p>\n<p>The research represents a crucial recognition that advanced AI in science demands more than technical innovation\u2014it requires a fundamental commitment to aligning powerful tools with human values and societal expectations.<\/p>\n<p><a href=\"https:\/\/doi.org\/10.1088\/3050-287X\/adfee5\">AI for Science: 10.1088\/3050-287X\/adfee5<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Scientists have created a digital gatekeeper that could prevent artificial intelligence from accidentally becoming a how-to manual for making dangerous chemicals, while still allowing legitimate research to flourish. The system, called SciGuard, acts like a sophisticated bouncer for AI models used in chemistry labs worldwide. When someone asks an AI system how to synthesize a &#8230; <a title=\"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress\" class=\"read-more\" href=\"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/\" aria-label=\"Read more about AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress\">Read more<\/a><\/p>\n","protected":false},"author":1299,"featured_media":266,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","_links_to":"","_links_to_target":""},"categories":[10,2],"tags":[],"class_list":["post-265","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-society","category-technology","generate-columns","tablet-grid-50","mobile-grid-100","grid-parent","grid-50"],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v27.3 (Yoast SEO v27.3) - https:\/\/yoast.com\/product\/yoast-seo-premium-wordpress\/ -->\n<title>AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress - SciChi<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress\" \/>\n<meta property=\"og:description\" content=\"Scientists have created a digital gatekeeper that could prevent artificial intelligence from accidentally becoming a how-to manual for making dangerous chemicals, while still allowing legitimate research to flourish. The system, called SciGuard, acts like a sophisticated bouncer for AI models used in chemistry labs worldwide. When someone asks an AI system how to synthesize a ... Read more\" \/>\n<meta property=\"og:url\" content=\"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/\" \/>\n<meta property=\"og:site_name\" content=\"SciChi\" \/>\n<meta property=\"article:published_time\" content=\"2025-09-24T13:22:01+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/09\/ai-safeguards.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"700\" \/>\n\t<meta property=\"og:image:height\" content=\"599\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"SciChi\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"SciChi\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/\"},\"author\":{\"name\":\"SciChi\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#\\\/schema\\\/person\\\/9974872362fae8e6096bd8c6637cf082\"},\"headline\":\"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress\",\"datePublished\":\"2025-09-24T13:22:01+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/\"},\"wordCount\":732,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/wp-content\\\/uploads\\\/sites\\\/16\\\/2025\\\/09\\\/ai-safeguards.jpg\",\"articleSection\":[\"Society\",\"Technology\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/#respond\"]}],\"copyrightYear\":\"2025\",\"copyrightHolder\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/#organization\"}},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/\",\"url\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/\",\"name\":\"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress - SciChi\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/wp-content\\\/uploads\\\/sites\\\/16\\\/2025\\\/09\\\/ai-safeguards.jpg\",\"datePublished\":\"2025-09-24T13:22:01+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/#primaryimage\",\"url\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/wp-content\\\/uploads\\\/sites\\\/16\\\/2025\\\/09\\\/ai-safeguards.jpg\",\"contentUrl\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/wp-content\\\/uploads\\\/sites\\\/16\\\/2025\\\/09\\\/ai-safeguards.jpg\",\"width\":700,\"height\":599,\"caption\":\"Overview of AI risks and SciGuard framework.\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/2025\\\/09\\\/24\\\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#website\",\"url\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/\",\"name\":\"SciChi\",\"description\":\"Tracking Chinese Research\",\"publisher\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#organization\",\"name\":\"SciChi\",\"url\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/wp-content\\\/uploads\\\/sites\\\/16\\\/2025\\\/04\\\/scichi-logo-cropped.jpg\",\"contentUrl\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/wp-content\\\/uploads\\\/sites\\\/16\\\/2025\\\/04\\\/scichi-logo-cropped.jpg\",\"width\":796,\"height\":296,\"caption\":\"SciChi\"},\"image\":{\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#\\\/schema\\\/logo\\\/image\\\/\"}},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/#\\\/schema\\\/person\\\/9974872362fae8e6096bd8c6637cf082\",\"name\":\"SciChi\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/45bfcb06f83fff507782e1030e14a31f738fce0220fc6a8fea863d633e61311f?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/45bfcb06f83fff507782e1030e14a31f738fce0220fc6a8fea863d633e61311f?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/45bfcb06f83fff507782e1030e14a31f738fce0220fc6a8fea863d633e61311f?s=96&d=mm&r=g\",\"caption\":\"SciChi\"},\"url\":\"https:\\\/\\\/scienceblog.com\\\/sciencechina\\\/author\\\/chinaresearch\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress - SciChi","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/","og_locale":"en_US","og_type":"article","og_title":"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress","og_description":"Scientists have created a digital gatekeeper that could prevent artificial intelligence from accidentally becoming a how-to manual for making dangerous chemicals, while still allowing legitimate research to flourish. The system, called SciGuard, acts like a sophisticated bouncer for AI models used in chemistry labs worldwide. When someone asks an AI system how to synthesize a ... Read more","og_url":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/","og_site_name":"SciChi","article_published_time":"2025-09-24T13:22:01+00:00","og_image":[{"width":700,"height":599,"url":"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/09\/ai-safeguards.jpg","type":"image\/jpeg"}],"author":"SciChi","twitter_card":"summary_large_image","twitter_misc":{"Written by":"SciChi","Est. reading time":"4 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/#article","isPartOf":{"@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/"},"author":{"name":"SciChi","@id":"https:\/\/scienceblog.com\/sciencechina\/#\/schema\/person\/9974872362fae8e6096bd8c6637cf082"},"headline":"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress","datePublished":"2025-09-24T13:22:01+00:00","mainEntityOfPage":{"@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/"},"wordCount":732,"commentCount":0,"publisher":{"@id":"https:\/\/scienceblog.com\/sciencechina\/#organization"},"image":{"@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/#primaryimage"},"thumbnailUrl":"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/09\/ai-safeguards.jpg","articleSection":["Society","Technology"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/#respond"]}],"copyrightYear":"2025","copyrightHolder":{"@id":"https:\/\/scienceblog.com\/#organization"}},{"@type":"WebPage","@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/","url":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/","name":"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress - SciChi","isPartOf":{"@id":"https:\/\/scienceblog.com\/sciencechina\/#website"},"primaryImageOfPage":{"@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/#primaryimage"},"image":{"@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/#primaryimage"},"thumbnailUrl":"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/09\/ai-safeguards.jpg","datePublished":"2025-09-24T13:22:01+00:00","breadcrumb":{"@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/#primaryimage","url":"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/09\/ai-safeguards.jpg","contentUrl":"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/09\/ai-safeguards.jpg","width":700,"height":599,"caption":"Overview of AI risks and SciGuard framework."},{"@type":"BreadcrumbList","@id":"https:\/\/scienceblog.com\/sciencechina\/2025\/09\/24\/ai-safety-net-blocks-toxic-chemical-recipes-while-preserving-scientific-progress\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/scienceblog.com\/sciencechina\/"},{"@type":"ListItem","position":2,"name":"AI Safety Net Blocks Toxic Chemical Recipes While Preserving Scientific Progress"}]},{"@type":"WebSite","@id":"https:\/\/scienceblog.com\/sciencechina\/#website","url":"https:\/\/scienceblog.com\/sciencechina\/","name":"SciChi","description":"Tracking Chinese Research","publisher":{"@id":"https:\/\/scienceblog.com\/sciencechina\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/scienceblog.com\/sciencechina\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/scienceblog.com\/sciencechina\/#organization","name":"SciChi","url":"https:\/\/scienceblog.com\/sciencechina\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/scienceblog.com\/sciencechina\/#\/schema\/logo\/image\/","url":"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/04\/scichi-logo-cropped.jpg","contentUrl":"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/04\/scichi-logo-cropped.jpg","width":796,"height":296,"caption":"SciChi"},"image":{"@id":"https:\/\/scienceblog.com\/sciencechina\/#\/schema\/logo\/image\/"}},{"@type":"Person","@id":"https:\/\/scienceblog.com\/sciencechina\/#\/schema\/person\/9974872362fae8e6096bd8c6637cf082","name":"SciChi","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/45bfcb06f83fff507782e1030e14a31f738fce0220fc6a8fea863d633e61311f?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/45bfcb06f83fff507782e1030e14a31f738fce0220fc6a8fea863d633e61311f?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/45bfcb06f83fff507782e1030e14a31f738fce0220fc6a8fea863d633e61311f?s=96&d=mm&r=g","caption":"SciChi"},"url":"https:\/\/scienceblog.com\/sciencechina\/author\/chinaresearch\/"}]}},"jetpack_featured_media_url":"https:\/\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/09\/ai-safeguards.jpg","jetpack_sharing_enabled":true,"jetpack_likes_enabled":true,"jetpack-related-posts":[{"id":381,"url":"https:\/\/scienceblog.com\/sciencechina\/2026\/03\/03\/three-ai-superpowers-are-developing-incompatible-technologies-that-may-never-converge\/","url_meta":{"origin":265,"position":0},"title":"Three AI Superpowers Are Developing Incompatible Technologies That May Never Converge","author":"SciChi","date":"March 3, 2026","format":false,"excerpt":"Somewhere in a Guangzhou laboratory, a team of researchers has just finished mapping something that looks, from a distance, like a geopolitical fault line \u2014 except it runs through lines of code rather than continental crust. The fracture they've been tracing separates three distinct technological civilisations: American, Chinese and European.\u2026","rel":"","context":"In &quot;Society&quot;","block_context":{"text":"Society","link":"https:\/\/scienceblog.com\/sciencechina\/category\/society\/"},"img":{"alt_text":"The AI triad: divergent technological pathways and their global implications","src":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2026\/03\/triad.jpeg?resize=350%2C200&ssl=1","width":350,"height":200,"srcset":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2026\/03\/triad.jpeg?resize=350%2C200&ssl=1 1x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2026\/03\/triad.jpeg?resize=525%2C300&ssl=1 1.5x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2026\/03\/triad.jpeg?resize=700%2C400&ssl=1 2x"},"classes":[]},{"id":47,"url":"https:\/\/scienceblog.com\/sciencechina\/2025\/04\/22\/new-ai-drones-spot-victims-through-forest-darkness\/","url_meta":{"origin":265,"position":1},"title":"New AI Drones Spot Victims Through Forest, Darkness","author":"SciChi","date":"April 22, 2025","format":false,"excerpt":"Finding survivors in wilderness after disasters has long been a race against time, with rescue teams battling challenging terrain and weather conditions. Now, researchers have developed an artificial intelligence system that dramatically improves how drones detect people during search and rescue missions\u2014even in dense forests or darkness. A team from\u2026","rel":"","context":"In &quot;Technology&quot;","block_context":{"text":"Technology","link":"https:\/\/scienceblog.com\/sciencechina\/category\/technology\/"},"img":{"alt_text":"The illustrated scenarios are based on locations in Xinyang and Huizhou, China. The system collected samples that included both RGB and infrared spectral bands, which were then used to construct the VTSaR dataset.","src":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/04\/remotesensing.0474.fig_.006.jpg?resize=350%2C200&ssl=1","width":350,"height":200,"srcset":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/04\/remotesensing.0474.fig_.006.jpg?resize=350%2C200&ssl=1 1x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/04\/remotesensing.0474.fig_.006.jpg?resize=525%2C300&ssl=1 1.5x"},"classes":[]},{"id":252,"url":"https:\/\/scienceblog.com\/sciencechina\/2025\/08\/25\/ai-breakthrough-reveals-hidden-storm-patterns-in-earths-atmosphere\/","url_meta":{"origin":265,"position":2},"title":"AI Breakthrough Reveals Hidden Storm Patterns in Earth&#8217;s Atmosphere","author":"SciChi","date":"August 25, 2025","format":false,"excerpt":"A powerful new AI tool is uncovering the fine details of storms that traditional weather models miss. Researchers from Wroc\u0142aw University of Environmental and Life Sciences, working with collaborators in Poland and California, have developed the first deep learning framework capable of generating high-resolution GNSS tomography of the atmosphere. Their\u2026","rel":"","context":"In &quot;Environment&quot;","block_context":{"text":"Environment","link":"https:\/\/scienceblog.com\/sciencechina\/category\/environment\/"},"img":{"alt_text":"clouds","src":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/08\/clouds-3488632_640.jpg?resize=350%2C200&ssl=1","width":350,"height":200,"srcset":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/08\/clouds-3488632_640.jpg?resize=350%2C200&ssl=1 1x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/08\/clouds-3488632_640.jpg?resize=525%2C300&ssl=1 1.5x"},"classes":[]},{"id":309,"url":"https:\/\/scienceblog.com\/sciencechina\/2025\/12\/08\/new-ai-reads-brainwaves-faster-by-mixing-neural-noise\/","url_meta":{"origin":265,"position":3},"title":"New AI Reads Brainwaves Faster by Mixing Neural Noise","author":"SciChi","date":"December 8, 2025","format":false,"excerpt":"Brain-computer interfaces have a frustrating problem. The deep learning models that power them are data gluttons, but getting clean brain recordings is expensive and slow. Every person's brain is different, which makes training these systems a nightmare. Now, researchers from Tianjin University and UC San Diego think they've found a\u2026","rel":"","context":"In &quot;Brain &amp; Behavior&quot;","block_context":{"text":"Brain &amp; Behavior","link":"https:\/\/scienceblog.com\/sciencechina\/category\/brain-behavior\/"},"img":{"alt_text":"flowchart of findings","src":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/12\/neural-noise.jpeg?resize=350%2C200&ssl=1","width":350,"height":200,"srcset":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/12\/neural-noise.jpeg?resize=350%2C200&ssl=1 1x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/12\/neural-noise.jpeg?resize=525%2C300&ssl=1 1.5x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/12\/neural-noise.jpeg?resize=700%2C400&ssl=1 2x"},"classes":[]},{"id":321,"url":"https:\/\/scienceblog.com\/sciencechina\/2025\/12\/28\/when-ai-writes-like-your-boss-trust-becomes-the-vulnerability\/","url_meta":{"origin":265,"position":4},"title":"When AI Writes Like Your Boss, Trust Becomes the Vulnerability","author":"SciChi","date":"December 28, 2025","format":false,"excerpt":"An email lands in your inbox. The tone is familiar, the phrasing matches your manager's style down to the occasional comma splice. It asks you to review a document on a shared drive. You click. By the time you realize the sender wasn't your boss, your password is already logged\u2026","rel":"","context":"In &quot;Society&quot;","block_context":{"text":"Society","link":"https:\/\/scienceblog.com\/sciencechina\/category\/society\/"},"img":{"alt_text":"abstract illustration of an email and trust","src":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/12\/unnamed-2.jpg?resize=350%2C200&ssl=1","width":350,"height":200,"srcset":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/12\/unnamed-2.jpg?resize=350%2C200&ssl=1 1x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/12\/unnamed-2.jpg?resize=525%2C300&ssl=1 1.5x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2025\/12\/unnamed-2.jpg?resize=700%2C400&ssl=1 2x"},"classes":[]},{"id":336,"url":"https:\/\/scienceblog.com\/sciencechina\/2026\/01\/04\/artificial-empathy-and-the-future-of-lonely-recovery\/","url_meta":{"origin":265,"position":5},"title":"Artificial Empathy and the Future of Lonely Recovery","author":"SciChi","date":"January 4, 2026","format":false,"excerpt":"Stroke patients grinding through repetitive arm exercises know the look: a therapist checking the clock, mentally calculating how many more patients need attention before shift end. That glance, however brief, changes the room. Healthcare systems worldwide are hemorrhaging staff faster than training programs can replace them, and researchers are asking\u2026","rel":"","context":"In &quot;Brain &amp; Behavior&quot;","block_context":{"text":"Brain &amp; Behavior","link":"https:\/\/scienceblog.com\/sciencechina\/category\/brain-behavior\/"},"img":{"alt_text":"robot handing a flower to human female","src":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2026\/01\/pexels-pavel-danilyuk-8438979-2.jpg?resize=350%2C200&ssl=1","width":350,"height":200,"srcset":"https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2026\/01\/pexels-pavel-danilyuk-8438979-2.jpg?resize=350%2C200&ssl=1 1x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2026\/01\/pexels-pavel-danilyuk-8438979-2.jpg?resize=525%2C300&ssl=1 1.5x, https:\/\/i0.wp.com\/scienceblog.com\/sciencechina\/wp-content\/uploads\/sites\/16\/2026\/01\/pexels-pavel-danilyuk-8438979-2.jpg?resize=700%2C400&ssl=1 2x"},"classes":[]}],"_links":{"self":[{"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/posts\/265","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/users\/1299"}],"replies":[{"embeddable":true,"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/comments?post=265"}],"version-history":[{"count":1,"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/posts\/265\/revisions"}],"predecessor-version":[{"id":267,"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/posts\/265\/revisions\/267"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/media\/266"}],"wp:attachment":[{"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/media?parent=265"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/categories?post=265"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/scienceblog.com\/sciencechina\/wp-json\/wp\/v2\/tags?post=265"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}