{"id":13544,"date":"2023-08-31T10:57:16","date_gmt":"2023-08-31T10:57:16","guid":{"rendered":"http:\/\/TheNextWeb=1399413"},"modified":"2023-08-31T10:57:16","modified_gmt":"2023-08-31T10:57:16","slug":"llms-prone-to-data-poisoning-and-prompt-injection-risks-uk-authority-warns","status":"publish","type":"post","link":"https:\/\/www.londonchiropracter.com\/?p=13544","title":{"rendered":"LLMs prone to data poisoning and prompt injection risks, UK authority warns"},"content":{"rendered":"\n<div><img decoding=\"async\" src=\"https:\/\/img-cdn.tnwcdn.com\/image\/tnw-blurple?filter_last=1&amp;fit=1280%2C640&amp;url=https%3A%2F%2Fcdn0.tnwcdn.com%2Fwp-content%2Fblogs.dir%2F1%2Ffiles%2F2023%2F08%2FUntitled-design-13-7.jpg&amp;signature=565a46f33b4d73517e0921cba118dd64\" class=\"ff-og-image-inserted\"><\/div>\n<p>The UK\u2019s National Cyber Security Centre (NCSC) is warning organisations to be wary of the imminent cyber risks associated with the integration of Large Language Models (LLMs) \u2014 such as ChatGPT \u2014 into their business, products, or services.<\/p>\n<p>In a <a href=\"https:\/\/www.ncsc.gov.uk\/blog-post\/exercise-caution-building-off-llms\" target=\"_blank\" rel=\"nofollow noopener\">set of<\/a> <a href=\"https:\/\/www.ncsc.gov.uk\/blog-post\/thinking-about-security-ai-systems\" target=\"_blank\" rel=\"nofollow noopener\">blog posts<\/a>, the NCSC emphasised that the global <a href=\"https:\/\/thenextweb.com\/topic\/tech\" target=\"_blank\" rel=\"noopener\">tech<\/a> community doesn\u2019t yet fully grasp LLMs\u2019 capabilities, weaknesses, and (most importantly) vulnerabilities. \u201cYou could say our understanding of LLMs is still \u2018in beta\u2019,\u2019\u2019 the authority said.<\/p>\n<p>One of the most extensively reported security weaknesses of existing LLMs is their susceptibility to malicious \u201cprompt injection\u201d attacks. These occur when a user creates an input aimed at causing the <a href=\"https:\/\/thenextweb.com\/topic\/ai\" target=\"_blank\" rel=\"noopener\">AI<\/a> model to behave in an unintended way \u2014 such as generating offensive content or disclosing confidential information.<\/p>\n<p>In addition, the data LLMs are trained on poses a twofold risk. Firstly a vast amount of this <a href=\"https:\/\/thenextweb.com\/topic\/data\" target=\"_blank\" rel=\"noopener\">data<\/a> is collected from the open internet, meaning it can include content that\u2019s inaccurate, controversial, or biased.<\/p>\n<div class=\"inarticle-wrapper channel-cta\">\n<div class=\"ica-text\" readability=\"0\"><a href=\"https:\/\/thenextweb.com\/events\/talks\" data-event-category=\"Article\" data-event-action=\"In Article Block\" data-event-label=\"Catch up on our conference talks\" target=\"_blank\" readability=\"4\" rel=\"noopener\"><\/p>\n<p class=\"ica-text__title\">Catch up on our conference talks<\/p>\n<p>Watch videos of our past talks for free with TNW All Access \u2192<\/p>\n<p><\/a><\/div>\n<\/div>\n<p>Secondly, cyber criminals can not only distort the data available for malicious practices (also known as \u201cdata poisoning\u201d), but also use it to conceal prompt injection attacks. This way, for example, a bank\u2019s AI-assistant for account holders can be tricked into transferring money to the attackers.<\/p>\n<p>\u201cThe emergence of LLMs is undoubtedly a very exciting time in technology \u2013 and a lot of people and organisations (including the NCSC) want to explore and benefit from it,\u201d said the authority.<\/p>\n<p>\u201cHowever, organisations building services that use LLMs need to be careful, in the same way they would be if they were using a product or code library that was in beta,\u201d the NCSC added. That is, with caution.<\/p>\n<p>The UK authority is urging organisations to establish cybersecurity principles and ensure that even the \u201cworst case scenario\u201d of whatever their LLM-powered applications are permitted to do is something they can deal with.<\/p>\n<p> <a href=\"https:\/\/thenextweb.com\/news\/llms-data-poisoning-prompt-injection-risks\">Source<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>The UK\u2019s National Cyber Security Centre (NCSC) is warning organisations to be wary of the imminent cyber risks associated with the integration of Large Language Models (LLMs) \u2014 such as ChatGPT \u2014&#8230;<\/p>\n","protected":false},"author":1,"featured_media":13545,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[1],"tags":[],"_links":{"self":[{"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=\/wp\/v2\/posts\/13544"}],"collection":[{"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=13544"}],"version-history":[{"count":0,"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=\/wp\/v2\/posts\/13544\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=\/wp\/v2\/media\/13545"}],"wp:attachment":[{"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=13544"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=13544"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.londonchiropracter.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=13544"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}