{"id":7726,"date":"2025-09-22T01:15:42","date_gmt":"2025-09-22T01:15:42","guid":{"rendered":"https:\/\/jobuzo.com\/en\/deepseek-warns-of-jailbreak-risks-for-its-open-source-models\/"},"modified":"2025-09-22T01:15:42","modified_gmt":"2025-09-22T01:15:42","slug":"deepseek-warns-of-jailbreak-risks-for-its-open-source-models","status":"publish","type":"post","link":"https:\/\/jobuzo.com\/en\/deepseek-warns-of-jailbreak-risks-for-its-open-source-models\/","title":{"rendered":"DeepSeek warns of \u2018jailbreak\u2019 risks for its open-source models"},"content":{"rendered":"<div>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">DeepSeek has revealed details about the risks posed by its artificial intelligence models for the first time, noting that open-sourced models are particularly susceptible to being &ldquo;jailbroken&rdquo; by malicious actors.<\/p>\n<div data-qa=\"InlineAdSlot-Container\" class=\"css-zl1inp e11v3ui14\">\n<div class=\"e11v3ui10 e11v3ui13 css-y2bwcc e1flwkbl0\" data-qa=\"AdSlot-Container\">\n<p>Advertisement<\/p>\n<\/div>\n<\/div>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">The Hangzhou-based start-up said it evaluated its models using industry benchmarks as well as its own tests in a peer-reviewed article published in the academic journal Nature.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">American AI companies often publicise research about the risks of their rapidly improving models and have introduced risk mitigation policies in response, such as Anthropic&rsquo;s Responsible Scaling Policies and OpenAI&rsquo;s Preparedness Framework.<\/p>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">Chinese companies were less outspoken about risks, despite their models being just a few months behind their US equivalents, according to AI experts. However, DeepSeek had conducted evaluations of such risks before, including the most serious &ldquo;frontier risks&rdquo;, the Post reported earlier.<\/p>\n<div class=\"methode-html-wrapper oembed-wrapper e1a5rv550 css-1llrc1m e1yqhwb40\" data-qa=\"Component-renderMap-StyledDiv\">\n<div>\n<div class=\"c1 e1drg7e30 e1ciypty0 ehdmpxk0 css-gm7mch e1hpd36k0\" data-qa=\"Component-SCMPYoutubeVideoContainer\" readability=\"6\">\n<div data-qa=\"SCMPYoutubeVideoPreview-PreviewContainer\" class=\"css-ksh7lk e1o401186\"><img decoding=\"async\" data-qa=\"BaseImage-handleRenderImage-StyledImage\" class=\"e1o401188 css-1w1l3op e445x7d0\" loading=\"lazy\" src=\"https:\/\/cdn.i-scmp.com\/sites\/default\/files\/styles\/wide_landscape\/public\/d8\/video\/thumbnail\/2025\/06\/10\/Thu_Clean-(15).jpg?itok=gnISwi22\">\n<div data-qa=\"SCMPYoutubeVideoPreview-PreviewContentContainer\" class=\"css-1giwmmb e1o401187\">\n<div data-qa=\"SCMPYoutubeVideoPreview-PreviewContentFloatContainer\" class=\"css-7nqkgi e1o401183\" readability=\"6\">\n<div data-qa=\"SCMPYoutubeVideoPreview-PreviewContentDataContainer\" class=\"css-1jla6kt e1o401180\" readability=\"7\">\n<p data-qa=\"SCMPYoutubeVideoPreview-PreviewDuration\" class=\"css-1vqvq42 e1o401181\">10:41<\/p>\n<div class=\"internal-linking-related-contents\"><a href=\"https:\/\/jobuzo.com\/en\/12-weeks-jail-for-school-it-support-technician-who-took-upskirt-videos-of-teachers\/\" class=\"template-1\"><span class=\"cta\">News :<\/span><span class=\"postTitle\">&lt;div&gt;12 weeks' jail for school IT support technician who took upskirt videos of teachers&lt;\/div&gt;<\/span><\/a><\/div><p data-qa=\"SCMPYoutubeVideoPreview-PreviewTitle\" class=\"css-zasw6y e1o401182\">How Hangzhou&rsquo;s &lsquo;Six Little Dragons&rsquo; built a new Chinese tech hub<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<div class=\"e1hpd36k1 css-79hwxn e1x11wau0\" data-qa=\"SCMPYouTubeVideoFooter-VideoTitleContainer\" readability=\"7\"><svg width=\"24\" height=\"25\" viewbox=\"0 0 24 25\" fill=\"none\" data-qa=\"SCMPYouTubeVideoFooter-StyledVideoRecorder\" class=\"css-1nnwg93 e1x11wau1\"><path fill-rule=\"evenodd\" clip-rule=\"evenodd\" d=\"M22 6.22926V18.4315L17.5463 13.9868V17.8655C17.5463 18.1618 17.4383 18.4248 17.2228 18.6531C17.0067 18.8829 16.7498 18.9967 16.4537 18.9967H3.13414C2.80989 18.9967 2.53968 18.8829 2.32425 18.6531C2.10808 18.4248 2 18.1618 2 17.8655V6.79519C2 6.49893 2.10808 6.23665 2.32425 6.00688C2.53968 5.77858 2.80989 5.66333 3.13414 5.66333H16.4537C16.7498 5.66333 17.0067 5.77858 17.2228 6.00688C17.4383 6.23665 17.5463 6.49893 17.5463 6.79519V10.674L22 6.22926Z\" fill=\"#001246\"><\/path><\/svg>\n<p>How Hangzhou&rsquo;s &lsquo;Six Little Dragons&rsquo; built a new Chinese tech hub<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<p datatype=\"p\" data-qa=\"Component-Component\" class=\"e8zc9q40 css-1c6uqr6 ec74h0k1\">The Nature paper provided more &ldquo;granular&rdquo; details about DeepSeek&rsquo;s testing regime, said Fang Liang, an expert member of China&rsquo;s AI Industry Alliance (AIIA), an industry body. These included &ldquo;red-team&rdquo; tests based on a framework introduced by Anthropic, in which testers try to get AI models to produce harmful speech.<\/p>\n<div data-qa=\"InlineAdSlot-Container\" class=\"css-zl1inp e11v3ui14\">\n<div class=\"e11v3ui10 e11v3ui13 css-gy323d e1flwkbl0\" data-qa=\"AdSlot-Container\">\n<p>Advertisement<\/p>\n<\/div>\n<\/div>\n<\/div>\n<p><sub>DeepSeek warns of &lsquo;jailbreak&rsquo; risks for its open-source models<\/sub><\/p>\n","protected":false},"excerpt":{"rendered":"<p>DeepSeek has revealed details about the risks posed by its artificial intelligence models for the first time, noting that open-sourced models are particularly susceptible to being &ldquo;jailbroken&rdquo; by malicious actors. Advertisement The Hangzhou-based start-up said it evaluated its models using industry benchmarks as well as its own tests in a peer-reviewed article published in the&#8230;<\/p>\n<p class=\"more-link-wrap\"><a href=\"https:\/\/jobuzo.com\/en\/deepseek-warns-of-jailbreak-risks-for-its-open-source-models\/\" class=\"more-link\">Read More<span class=\"screen-reader-text\"> &ldquo;DeepSeek warns of \u2018jailbreak\u2019 risks for its open-source models&rdquo;<\/span> &raquo;<\/a><\/p>\n","protected":false},"author":1,"featured_media":7727,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-7726","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news"],"_links":{"self":[{"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/posts\/7726","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/comments?post=7726"}],"version-history":[{"count":0,"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/posts\/7726\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/media\/7727"}],"wp:attachment":[{"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/media?parent=7726"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/categories?post=7726"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/jobuzo.com\/en\/wp-json\/wp\/v2\/tags?post=7726"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}