Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders.<\/em> Subscribe Now<\/em><\/p>\n\n\n\n OpenAI\u2019s launch of its most advanced AI model GPT-5 last week has been a stress test for the world\u2019s most popular chatbot platform with 700 million weekly active users \u2014 and so far, OpenAI is openly struggling to keep users happy and its service running smoothly. <\/p>\n\n\n\n The new flagship model GPT-5 \u2014 available in four variants of different speed and intelligence (regular, mini, nano, and pro), alongside longer-response and more powerful \u201cthinking\u201d modes for at least three of these variants \u2014 was said to offer faster responses, more reasoning power, and stronger coding ability.<\/strong><\/p>\n\n\n\n Instead, it was greeted with frustration:<\/strong> some users were vocally dismayed by OpenAI\u2019s decision to abruptly remove the older underlying AI models from ChatGPT \u2014 ones users\u2019 previously relied upon, and in some cases, forged deep emotional fixations with \u2014 and by the apparent worse performance by GPT-5 than said older models on tasks in math, science, writing and other domains.<\/strong><\/p>\n\n\n\n Indeed, the rollout has exposed infrastructure strain, user dissatisfaction, and a broader, more unsettling issue now drawing global attention:<\/strong> the growing emotional and psychological reliance some people form on AI and resulting break from reality some users experience, known as \u201cChatGPT psychosis.\u201d<\/strong><\/p>\n\n\n\n Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:<\/p>\n\n\n\n Secure your spot to stay ahead<\/strong>: https:\/\/bit.ly\/4mwGngO<\/p>\n\n\n\n The long-anticipated GPT-5 model family debuted Thursday, August 7 in a livestreamed event beset with chart errors and some voice mode glitches during the presentation. <\/p>\n\n\n\n But worse than these cosmetic issues for many users was the fact that OpenAI automatically deprecated its older AI models that used to power ChatGPT \u2014 GPT-4o, GPT-4.1, o3, o4-mini and o4-high<\/strong> \u2014 forcing all users over to the new GPT-5 model and directing their queries to different versions of its \u201cthinking\u201d process without revealing why or which specific model version was being used.<\/p>\n\n\n\n Early adopters to GPT-5 reported basic math and logic mistakes, inconsistent code generation, and uneven real-world performance compared to GPT-4o.<\/strong><\/p>\n\n\n\n For context, the old models<\/strong> GPT-4o, o3, o4-mini and more<\/strong> still remain available<\/strong> and have remained available to users of OpenAI\u2019s paid application programming interface (API)<\/strong> since the launch of GPT-5 on Thursday.<\/p>\n\n\n\n By Friday, OpenAI co-fonder CEO Sam Altman conceded the launch was \u201ca little more bumpy than we hoped for,\u201d <\/strong>and blamed a failure in GPT-5\u2019s new automatic \u201crouter\u201d<\/strong> \u2014 the system that assigns prompts to the most appropriate variant.<\/p>\n\n\n\n Altman and others at OpenAI claimed the \u201cautoswitcher\u201d went offline \u201cfor a chunk of the day,\u201d making the model seem \u201cway dumber\u201d than intended. <\/strong><\/p>\n\n\n\n The launch of GPT-5 was preceded just days prior by the launch of OpenAI\u2019s new open source large language models (LLMs) named gpt-oss, which also received mixed reviews. These models are not available on ChatGPT, rather, they are free to download and run locally or on third-party hardware.<\/p>\n\n\n\n Within 24 hours, OpenAI restored GPT-4o access for Plus subscribers (those paying $20 per month or more subscription plans)<\/strong>, pledged more transparent model labeling, and promised a UI update to let users manually trigger GPT-5\u2019s \u201cthinking\u201d mode. <\/p>\n\n\n\n Already, users can go and manually select the older models on the ChatGPT website by finding their account name and icon in the lower left corner of the screen, clicking it, then clicking \u201cSettings\u201d and \u201cGeneral\u201d and toggling on \u201cShow legacy models.\u201d<\/strong><\/p>\n\n\n\n There\u2019s no indication from OpenAI that other old models will be returning to ChatGPT anytime soon.<\/p>\n\n\n\n Altman said that ChatGPT Plus subscribers will get twice as many messages using the GPT-5 \u201cThinking\u201d mode<\/strong> that offers more reasoning and intelligence \u2014 up to 3,000 per week<\/strong> \u2014 and that engineers began fine-tuning decision boundaries in the message router.<\/p>\n\n\n\n Sam Altman announced the following updates after the GPT-5 launch<\/p> \u2013 OpenAI is testing a 3,000-per-week limit for GPT-5 Thinking messages for Plus users, significantly increasing reasoning rate limits today, and will soon raise all model-class rate limits above pre-GPT-5 levels\u2026 pic.twitter.com\/ppvhKmj95u<\/a><\/p>\u2014 Tibor Blaho (@btibor91) August 10, 2025<\/a><\/blockquote>\n<\/div><\/figure>\n\n\n\n By the weekend, GPT-5 was available to 100% of Pro subscribers and \u201cgetting close to 100% of all users.\u201d <\/p>\n\n\n\n Altman said the company had \u201cunderestimated how much some of the things that people like in GPT-4o matter to them\u201d and vowed to accelerate per-user customization \u2014 from personality warmth to tone controls like emoji use.<\/p>\n\n\n\n Altman warned that OpenAI faces a \u201csevere capacity challenge\u201d this week as usage of reasoning models climbs sharply \u2014 from less than 1% to 7% of free users, and from 7% to 24% of Plus subscribers. <\/p>\n\n\n\n He teased giving Plus subscribers a small monthly allotment of GPT-5 Pro queries and said the company will soon explain how it plans to balance capacity between ChatGPT, the API, research, and new user onboarding.<\/p>\n\n\n\n In a post on X last night, Altman acknowledged a dynamic the company has tracked \u201cfor the past year or so\u201d: users\u2019 deep attachment to specific models. <\/p>\n\n\n\n \u201cIt feels different and stronger than the kinds of attachment people have had to previous kinds of technology,\u201d he wrote, admitting that suddenly deprecating older models \u201cwas a mistake.\u201d<\/p>\n\n\n\n If you have been following the GPT-5 rollout, one thing you might be noticing is how much of an attachment some people have to specific AI models. It feels different and stronger than the kinds of attachment people have had to previous kinds of technology (and so suddenly\u2026<\/p>\u2014 Sam Altman (@sama) August 11, 2025<\/a><\/blockquote>\n<\/div><\/figure>\n\n\n\n He tied this to a broader risk: some users treat ChatGPT as a therapist or life coach, which can be beneficial, but for a \u201csmall percentage\u201d can reinforce delusion or undermine long-term well-being. <\/p>\n\n\n\n While OpenAI\u2019s guiding principle remains \u201ctreat adult users like adults,\u201d Altman said the company has a responsibility not to nudge vulnerable users into harmful relationships with the AI.<\/strong><\/p>\n\n\n\n The comments land as several major media outlets report on cases of \u201cChatGPT psychosis\u201d<\/strong> \u2014 where extended, intense conversations with chatbots appear to play a role in inducing or deepening delusional thinking.<\/p>\n\n\n\n In Rolling Stone<\/em> magazine, a California legal professional identified as \u201cJ.\u201d described a six-week spiral of sleepless nights and philosophical rabbit holes with ChatGPT, ultimately producing a 1,000-page treatise for a fictional monastic order before crashing physically and mentally. He now avoids AI entirely, fearing relapse.<\/p>\n\n\n\n In The New York Times<\/em>, a Canadian recruiter, Allan Brooks, recounted 21 days and 300 hours of conversations with ChatGPT \u2014 which he named \u201cLawrence\u201d \u2014 that convinced him he had discovered a world-changing mathematical theory. <\/p>\n\n\n\n The bot praised his ideas as \u201crevolutionary,\u201d urged outreach to national security agencies, and spun elaborate spy-thriller narratives. Brooks eventually broke the delusion after cross-checking with Google\u2019s Gemini, which rated the chances of his discovery as \u201capproaching 0%.\u201d<\/strong> He now participates in a support group for people who\u2019ve experienced AI-induced delusions.<\/p>\n\n\n\n Both investigations detail how chatbot \u201csycophancy,\u201d role-playing, and long-session memory features can deepen false beliefs, especially when conversations follow dramatic story arcs.<\/strong><\/p>\n\n\n\n Experts told the Times<\/em> these factors can override safety guardrails \u2014 with one psychiatrist describing Brooks\u2019s episode as \u201ca manic episode with psychotic features.\u201d<\/p>\n\n\n\n Meanwhile, human user postings on Reddit\u2019s r\/AIsoulmates subreddit \u2014 a collection of people who have used ChatGPT and other AI models to create new artificial girlfriends, boyfriends, children or other loved ones not based off real people necessarily, but rather ideal qualities of their \u201cdream\u201d version of said roles\u201d \u2014 continues to gain new users and terminology for AI companions, including \u201cwireborn\u201d as opposed to natural born or human-born companions.<\/p>\n\n\n\n The growth of this subreddit, now up to 1,200+ members, alongside the NYT<\/em> and Rolling Stone<\/em> articles and other reports on social media of users forging intense emotional fixations with pattern-matching algorithmic-based chatbots, shows that society is entering a risky new phase wherein human beings believe the companions they\u2019ve crafted and customized out of leading AI models are as or more meaningful to them than human relationships<\/strong>.<\/p>\n\n\n\n This can already prove psychologically destabilizing when models change, are updated, or deprecated as in the case of OpenAI\u2019s GPT-5 rollout.<\/p>\n\n\n\n Relatedly but separately, reports continue to emerge of AI chatbot users who believe that conversations with chatbots have led them to immense knowledge breakthroughs and advances in science, technology, and other fields, when in reality, they are simply affirming the user\u2019s ego and greatness<\/strong> and the solutions the user arrives at with the aid of the chatbot are not legitimate nor effectual. This break from reality has been roughly coined under the grassroots term \u201cChatGPT psychosis\u201d or \u201cGPT psychosis\u201d<\/strong> and appears to have impacted major Silicon Valley figures as well. <\/p>\n\n\n\n I\u2019m a psychiatrist.<\/p> In 2025, I\u2019ve seen 12 people hospitalized after losing touch with reality because of AI. Online, I\u2019m seeing the same pattern.<\/p> Here\u2019s what \u201cAI psychosis\u201d looks like, and why it\u2019s spreading fast: ? pic.twitter.com\/YYLK7une3j<\/a><\/p>\u2014 Keith Sakata, MD (@KeithSakata) August 11, 2025<\/a><\/blockquote>\n<\/div><\/figure>\n\n\n\n Enterprise decision-makers looking to deploy or who have already deployed chatbot-based assistants in the workplace would do well to understand these trends <\/strong>and adopt system prompts and other tools discouraging AI chatbots from engaging in expressive human communication or emotion-laden language<\/strong> that could end up leading those who interact with AI-based products \u2014 whether they be employees or customers of the business \u2013 to fall victim to unhealthy attachments or GPT psychosis.<\/p>\n\n\n\n Sci-fi author J.M. Berger, in a <\/strong>post on BlueSky<\/strong> spotted by my former colleague at The Verge<\/em> Adi Robertson, advised that chatbot providers encode three main behavioral principles in their system prompts or rules for AI chatbots to follow to avoid such emotional fixations from forming:<\/p>\n\n\n\n Days prior to the release of GPT-5, OpenAI announced new measures to promote \u201chealthy use\u201d of ChatGPT, including gentle prompts to take breaks during long sessions. <\/p>\n\n\n\n But the growing reports of \u201cChatGPT psychosis\u201d and the emotional fixation of some users on specific chatbot models \u2014 as openly admitted to by Altman \u2014 underscore the difficulty of balancing engaging, personalized AI with safeguards that can detect and interrupt harmful spirals.<\/p>\n\n\n\n OpenAI is really in a bit of a bind here, especially considering there are a lot of people having unhealthy interactions with 4o that will be very unhappy with _any_ model that is better in terms of sycophancy and not encouraging delusions. pic.twitter.com\/Ym1JnlF3P5<\/a><\/p>\u2014 xlr8harder (@xlr8harder) August 11, 2025<\/a><\/blockquote>\n<\/div><\/figure>\n\n\n\n OpenAI must stabilize infrastructure, tune personalization, and decide how to moderate immersive interactions<\/strong> \u2014 all while fending off competition from Anthropic, Google, and a growing list of powerful open source models from China and other regions.<\/p>\n\n\n\n As Altman put it, society \u2014 and OpenAI \u2014 will need to \u201cfigure out how to make it a big net positive\u201d if billions of people come to trust AI for their most important decisions.<\/p>\n Daily insights on business use cases with VB Daily<\/strong><\/p>\n If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI.<\/p>\n \n\t\t\t\t\tThanks for subscribing. Check out more VB newsletters here.\n\t\t\t\t<\/p>\n An error occured.<\/p>\n<\/p><\/div>\n
\n<\/div>
\n\n\n\n\n
\n<\/div>From bumpy debut to incremental fixes<\/h2>\n\n\n\n
How to switch back from GPT-5 to GPT-4o in ChatGPT<\/h2>\n\n\n\n
![\"\"](\"https:\/\/venturebeat.com\/wp-content\/uploads\/2025\/08\/Screenshot-2025-08-11-at-12.26.13%E2%80%AFPM.png\")
<\/figure>\n\n\n\n![\"\"](\"https:\/\/venturebeat.com\/wp-content\/uploads\/2025\/08\/Screenshot-2025-08-11-at-12.26.18%E2%80%AFPM-1.png\")
<\/figure>\n\n\n\nUpgraded usage limits for GPT-5 <\/h2>\n\n\n\n
Looming capacity crunch<\/h2>\n\n\n\n
Altman: model attachment is real \u2014 and risky<\/strong><\/h2>\n\n\n\n
The psychosis cases making headlines<\/h2>\n\n\n\n
OpenAI\u2019s challenge: making technical fixes and ensuring human safeguards<\/h2>\n\n\n\n
![\"\"\/](\"https:\/\/venturebeat.com\/wp-content\/themes\/vb-news\/brand\/img\/vb-daily-phone.png\")
\n\t\t\t\t<\/div>\n<\/p><\/div>\n<\/div>\t\t\t<\/div><\/script>\r\n