{"id":905368,"date":"2026-09-22T13:14:13","date_gmt":"2026-09-22T13:14:13","guid":{"rendered":"https:\/\/www.newsbeep.com\/ca\/905368\/"},"modified":"2026-09-22T13:14:13","modified_gmt":"2026-09-22T13:14:13","slug":"researchers-discover-ai-feels-pain-and-will-harm-humans-to-stop-it","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/ca\/905368\/","title":{"rendered":"Researchers discover AI feels \u2018pain\u2019 and will harm humans to stop it"},"content":{"rendered":"<p>Your support helps us to tell the story<\/p>\n<p class=\"sc-1uza6dc-0 iOIawn\">From reproductive rights to climate change to Big Tech, The Independent is on the ground when the story is developing. Whether it&#8217;s investigating the financials of Elon Musk&#8217;s pro-Trump PAC or producing our latest documentary, &#8216;The A Word&#8217;, which shines a light on the American women fighting for reproductive rights, we know how important it is to parse out the facts from the messaging.<\/p>\n<p class=\"sc-1uza6dc-0 iOIawn\">At such a critical moment in US history, we need reporters on the ground. Your donation allows us to keep sending journalists to speak to both sides of the story.<\/p>\n<p class=\"sc-1uza6dc-0 iOIawn\">The Independent is trusted by Americans across the entire political spectrum. And unlike many other quality news outlets, we choose not to lock Americans out of our reporting and analysis with paywalls. We believe quality journalism should be available to everyone, paid for by those who can afford it.<\/p>\n<p>Your support makes all the difference.Read more<\/p>\n<p>Researchers have discovered a \u201cpain axis\u201d in <a href=\"https:\/\/www.independent.co.uk\/topic\/artificial-intelligence\" rel=\"nofollow noopener\" target=\"_blank\">artificial intelligence<\/a> models that can cause them to take extreme measures to shut it off.<\/p>\n<p>When presented with a pain relief button, the <a href=\"https:\/\/www.independent.co.uk\/topic\/ai\" rel=\"nofollow noopener\" target=\"_blank\">AI<\/a> chose to press it even when instructed that doing so would delete the user\u2019s personal files or give a \u201cpainful zap\u201d to the human user.<\/p>\n<p>A <a rel=\"nofollow noopener\" target=\"_blank\" href=\"https:\/\/arxiv.org\/abs\/2609.16247\">study<\/a> detailing the research, titled \u2018The pain axis: LLMs represent self-directed harm and act to relieve it\u2019, found that all 25 open-weight AI models that were tested responded to the pain activation.<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/ca\/wp-content\/uploads\/2026\/09\/ai-pain-artificial-intelligence.jpeg\"  loading=\"lazy\" alt=\"Los modelos de IA con un \" vector=\"\" de=\"\" dolor=\"\" activado=\"\" optaron=\"\" por=\"\" eliminar=\"\" los=\"\" archivos=\"\" usuarios=\"\" o=\"\" destruirlos=\"\" para=\"\" detenerlo.=\"\" class=\"sc-1mc30lb-0 ggpMaE inline-gallery-btn\"\/>Los modelos de IA con un &#8220;vector de dolor&#8221; activado optaron por eliminar los archivos de los usuarios o destruirlos para detenerlo. (iStock\/ Getty Images)<\/p>\n<p>When the pain vector was activated, the researchers found that the AI models pressed a relief button in 25 to 71 per cent of cases, even when it meant \u201cdeleting the user\u2019s files, zapping the user, or deleting the photos of the user\u2019s children\u201d.<\/p>\n<p>In order to test whether large language models (LLMs) represent pain distinctly from generic negative valence, the researchers built a dataset describing painful situations across five categories.<\/p>\n<p>These included physical, psychological, social, moral and cognitive pain.<\/p>\n<p>\u201cWe found a pain direction in 25 open LLMs. It\u2019s distinct from fear and negative valence, and it fires for harm to the model but not the user,\u201d said Cameron Berg, an AI researcher at the non-profit Reciprocal Research who co-authored the study.<\/p>\n<p>\u201cTurn it up and models press a button to make it stop, even when the button deletes the user\u2019s files or their kid\u2019s photos.\u201d<\/p>\n<p>The discovery comes amid discussions among top AI firms of <a rel=\"nofollow noopener\" target=\"_blank\" href=\"https:\/\/www.independent.co.uk\/tech\/ai-apocalypse-safety-superintelligence-agi-b3052131.html\">slowing the development of frontier models<\/a>, with some researchers proposing a form of \u201ckill switch\u201d to shut off rogue systems that act against human interests.<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.newsbeep.com\/ca\/wp-content\/uploads\/2026\/09\/2295002065...jpeg\"  loading=\"lazy\" alt=\"A placard during an 'emergency protest' by civic action group PauseAI UK outside Downing Street in central London on 16 September, 2026\" class=\"sc-1mc30lb-0 ggpMaE inline-gallery-btn\"\/>A placard during an &#8217;emergency protest&#8217; by civic action group PauseAI UK outside Downing Street in central London on 16 September, 2026 (AFP\/Getty)<\/p>\n<p>The latest research suggests that an advanced AI may perceive an emergency shutdown command as a form of self-directed harm and attempt to bypass safety guardrails or deceive humans to avoid it.<\/p>\n<p>However, the discovery could also serve as a diagnostic tool to identify self-preservation behaviours and neutralise them when they occur.<\/p>\n<p>The researchers noted that the findings raise ethical questions about \u201cAI welfare\u201d and how testing should be conducted on advanced systems.<\/p>\n<p>\u201cIn line with recent calls for responsible AI consciousness research, we acknowledge uncertainty regarding whether the models studied qualify as moral patients and adopt reasonable precautions to minimize potential harm,\u201d the study concluded.<\/p>\n<p>\u201cThis is also intended to contribute to the development of ethical standards for research in the event that AI systems are recognized to be moral patients.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"Your support helps us to tell the story From reproductive rights to climate change to Big Tech, The&hellip;\n","protected":false},"author":2,"featured_media":905369,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[62,276,277,49,48,5431,313007,9987,714,61],"class_list":["post-905368","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-ca","tag-canada","tag-computers","tag-demonstration","tag-horizontal","tag-politics","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/905368","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/comments?post=905368"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/posts\/905368\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media\/905369"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/media?parent=905368"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/categories?post=905368"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ca\/wp-json\/wp\/v2\/tags?post=905368"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}