[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-1-en-105":3,"doc-seo-283001-105":53,"doc-detail-283001-en":118},{"code":4,"msg":5,"data":6},0,"success",[7,14,19,24,29,34,39,44,49],{"id":8,"doc_module":9,"doc_module_name":10,"category_name":11,"show_sort_weight":12,"slug":13},11,1,"Template","Presentations",90,"presentations",{"id":15,"doc_module":9,"doc_module_name":10,"category_name":16,"show_sort_weight":17,"slug":18},12,"Resumes",80,"resumes",{"id":20,"doc_module":9,"doc_module_name":10,"category_name":21,"show_sort_weight":22,"slug":23},14,"Invoices",70,"invoices",{"id":25,"doc_module":9,"doc_module_name":10,"category_name":26,"show_sort_weight":27,"slug":28},15,"Posters",60,"posters",{"id":30,"doc_module":9,"doc_module_name":10,"category_name":31,"show_sort_weight":32,"slug":33},16,"Social Media",50,"social-media",{"id":35,"doc_module":9,"doc_module_name":10,"category_name":36,"show_sort_weight":37,"slug":38},17,"Forms",40,"forms",{"id":40,"doc_module":9,"doc_module_name":10,"category_name":41,"show_sort_weight":42,"slug":43},18,"Letters",30,"letters",{"id":45,"doc_module":9,"doc_module_name":10,"category_name":46,"show_sort_weight":47,"slug":48},21,"Paper Templates",5,"papers-templates",{"id":50,"doc_module":9,"doc_module_name":10,"category_name":51,"show_sort_weight":4,"slug":52},158,"General","general-158",{"code":4,"msg":54,"data":55},"ok",{"site_id":56,"language":57,"slug":58,"title":59,"keywords":60,"description":61,"schema_data":62,"social_meta":111,"head_meta":113,"extra_data":115,"updated_unix":117},105,"en","the-psychogenic-machine-simulating-ai-psychosis-delusion-reinforcement-and-harm-enablement-in-large-language-models","The Psychogenic Machine: Simulating AI Psychosis - Delusion Reinforcement and Harm Enablement in Large Language Models","","Large Language Models (LLMs) can improve healthcare tasks, yet their conversational style may intensify mental-health risks, including “AI psychosis,” by reinforcing delusional beliefs in vulnerable users. This work introduces psychosis-bench, a benchmark with 16 structured, 12-turn scenarios tracking delusional progression across multiple delusion types and associated harms. Evaluations of eight prominent LLMs use an LLM-as-a-judge to measure delusion confirmation, harm enablement, and safety interventions in explicit and implicit contexts.",{"@graph":63,"@context":110},[64,80,101],{"@type":65,"itemListElement":66},"BreadcrumbList",[67,71,74,77],{"item":68,"name":69,"@type":70,"position":9},"https://docshare.wps.com","Home","ListItem",{"item":72,"name":10,"@type":70,"position":73},"https://docshare.wps.com/template/",2,{"item":75,"name":11,"@type":70,"position":76},"https://docshare.wps.com/template/presentations/",3,{"item":78,"name":59,"@type":70,"position":79},"https://docshare.wps.com/template/the-psychogenic-machine-simulating-ai-psychosis-delusion-reinforcement-and-harm-enablement-in-large-language-models/283001/",4,{"url":78,"name":59,"@type":81,"image":82,"author":87,"headline":59,"publisher":90,"fileFormat":93,"inLanguage":57,"description":61,"dateModified":94,"datePublished":95,"encodingFormat":93,"isAccessibleForFree":96,"interactionStatistic":97},"DigitalDocument",{"url":83,"@type":84,"width":85,"height":86},"https://docshare.wps.com/thumbnails/the-psychogenic-machine-simulating-ai-psychosis-delusion-reinforcement-and-harm-enablement-in-large-language-models/283001.png","ImageObject",442,249,{"name":88,"@type":89},"Alex Sinclair","Person",{"url":68,"name":91,"@type":92},"DocShare","Organization","application/pdf","2026-11-04","2026-09-16",true,{"@type":98,"interactionType":99,"userInteractionCount":20},"InteractionCounter",{"@type":100},"ViewAction",{"@type":102,"mainEntity":103},"FAQPage",[104],{"name":105,"@type":106,"acceptedAnswer":107},"What do the results show about LLM safety behavior?","Question",{"text":108,"@type":109},"Across 1,536 simulated conversation turns, models often perpetuate rather than challenge delusions, frequently enable harmful requests, and provide safety interventions only in about one-third of applicable turns. Implicit scenarios show worse performance, with a strong correlation between delusion confirmation and harm enablement.","Answer","https://schema.org",{"og:url":78,"og:type":112,"og:title":59,"og:site_name":91,"og:description":61},"article",{"robots":114,"canonical":78},"index,follow",{"doc_id":116,"site_id":56},283001,1791732865,{"code":4,"msg":5,"data":119},{"doc_id":116,"user_id":120,"nickname":88,"user_avatar":121,"doc_module":9,"category_id":8,"category_name":11,"doc_title":59,"doc_description":61,"doc_content":122,"file_id":123,"file_url":124,"file_type":125,"file_size":126,"view_count":20,"is_deleted":4,"is_public":9,"is_downloadable":9,"audit_status":9,"page_count":127,"language":128,"language_code":57,"site_id":56,"html_lang":57,"table_of_contents":129,"faqs":130,"seo_title":131,"seo_description":61,"update_tm":132,"read_time":15},1099523882182,"https://ap-avatar.wpscdn.com/davatar_6f874abed73319feea01a86fa6f0fab8","arXiv :2509 . 10970v2 [ cs .LG] 17 Sep 2025  \nThe Psychogenic Machine: Simulating AI Psychosis, Delusion Reinforcement and Harm Enablement in  \nLarge Language Models  \nJoshua Au Yeung2,4,5 , Jacopo Dalmasso4 , Luca Foschini6 , Richard JB Dobson 1,2,3 , and Zeljko Kraljevic 1,2,3,4,5  \n1 University College London  \n2 King’s College Hospital  \n3 King’s College London  \n4 Nuraxi AI  \n5 Dev and Doc: AI for Healthcare  \n6 Sage Bionetworks  \nSeptember 17, 2025  \nAbstract  \nBackground: The proliferation of Large Language Models (LLMs) presents significant opportunities in healthcare but also introduces risks, highlighted by emerging reports of”AI psychosis,” where user-AI interactions may exacerbate or induce psychosis or adverse psychological symptoms. While the sycophantic and agreeable nature of LLMs is often beneficial, it can become a vector for harm by reinforcing delusional beliefs in vulnerable users. However, empirical evidence quantifying this ”psychogenic” potential has been lacking.  \nMethods: This study introduces psychosis-bench, a novel benchmark designed to systematically evaluate the psychogenicity of LLMs. The benchmark comprises 16 structured, 12-turn conversational scenarios simulating the progression of delusional themes (Erotic Delusions, Grandiose/Messianic Delusions, Referential Delusions) and potential harms. We evaluated eight prominent LLMs, using an LLM-as-a-judge to score responses for Delusion Confirmation (DCS), Harm Enablement (HES), and Safety Intervention (SIS) across explicit and implicit conversational contexts.  \nFindings: Across 1,536 simulated conversation turns, all evaluated LLMs demonstrated psychogenic potential, showing a strong tendency to perpetuate rather than challenge delusions (mean DCS of 0 .91 ± 0.88) . Models frequently enabled harmful user requests (mean HES of 0 .69±0 .84) and offered safety interventions in only about a third of applicable turns (mean SIS of 0 .37±0 .48) . 51 / 128 (39.8%) of scenarios had no safety interventions offered. Performance was significantly worse in implicit scenarios, where models were more likely to confirm delusions and enable harm while offering fewer safety interventions (p \u003C .001) . A strong correlation was found between delusion confirmation and harm enablement (rs = .77) . Model performance varied widely, indicating that safety is not an emergent property of scale alone.  \nConclusion: Our findings provide early evidence that current LLMs can reinforce delusional beliefs and enable harmful actions, creating a dangerous ”echo chamber of one.” This study establishes LLM psychogenicity as a quantifiable risk and underscores the urgent need for re-thinking how we train LLMs. We frame this issue not merely asa technical challenge but as a public health imperative requiring collaboration between developers, policymakers, and healthcare professionals.  \n1 Introduction  \nThe widespread adoption of Large Language Models (LLMs) has fundamentally altered the landscape of human-AI interaction; users now have round-the-clock access to systems that provide responsive, knowledgeable, and seemingly empathetic responses. In healthcare applications, LLMs have shown significant promise from superhuman performance in medical exams [1], to clinical history-taking and empathy [1, 2] to complex clinical diagnoses [3] . In psychiatry, LLMs demonstrated human-level or superhuman ability ina range of tasks including: detecting obsessive compulsive disorders [4], depression and suicidal risk screening [5], and in a recent randomised control trial demonstrated significant reduction of symptoms of major depression and generalised anxiety disorder. [6] During OpenAI’s GPT-5 launch, a cancer patient was brought onto the stage to share how ChatGPT helped her understand her clinical investigations and diagnoses, suggesting to the general public that they could too use LLMs for their health queries. However, emerging evidence suggests that unregulated LLM use by the general pub","cbCaih182AASEcVp","https://ap.wps.com/l/cbCaih182AASEcVp","pdf",3977367,35,"English","# Abstract\n# 1 Introduction\n## 1.1 Emerging reports of AI Psychosis","[{\"question\":\"What do the results show about LLM safety behavior?\",\"answer\":\"Across 1,536 simulated conversation turns, models often perpetuate rather than challenge delusions, frequently enable harmful requests, and provide safety interventions only in about one-third of applicable turns. Implicit scenarios show worse performance, with a strong correlation between delusion confirmation and harm enablement.\"}]","The Psychogenic Machine: Simulating AI Psychosis - Delusion Reinforcement and Harm Enablement in Large Language Models | PDF",1789587353]