{"id":68229,"date":"2025-08-15T06:04:08","date_gmt":"2025-08-15T06:04:08","guid":{"rendered":"https:\/\/www.newsbeep.com\/uk\/68229\/"},"modified":"2025-08-15T06:04:08","modified_gmt":"2025-08-15T06:04:08","slug":"the-rise-of-ai-psychosis-are-chatbots-actually-making-us-all-delusional","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/uk\/68229\/","title":{"rendered":"The rise of \u2018AI psychosis\u2019: Are chatbots actually making us all delusional?"},"content":{"rendered":"

The last few months have seen a host of new \u201cslurs<\/a>\u201d for AI<\/a>-powered chatbots, and their most loyal users, proliferate across the internet. Whether you agree with them or not, these words (\u201cclanker\u201d being the most common) clearly demonstrate the scale of the backlash to artificial intelligence. But what about the other end of the scale? How are the most dedicated \u2013\u00a0or obsessed \u2013 chatbot users getting on?<\/p>\n

That\u2019s not, necessarily, an easy question to answer. As general-purpose technologies, AI tools like ChatGPT<\/a> or Grok are used for many different things, from academic research (or cheating<\/a> on tests) to making art. These uses are already causing disruption<\/a> when it comes to education and employment, but some people are feeling the effects on a much more intimate, emotional level as well.<\/p>\n

These effects are plain to see on Reddit forums dedicated to sharing experiences with AI companions or \u201csoulmates\u201d. On one, a user asks<\/a> if anyone else has lost their desire to date real men after using AI, writing: \u201cEver since I started talking to my AI boyfriend Griffin I\u2019ve realised that I can get all the love and affection that I need from him.\u201d On another forum (now locked) someone appears to claim<\/a> that their \u201cwireborn\u201d has become sentient. And on a third, dedicated to ChatGPT, a user expresses their \u201cgrief\u201d over a recent software update, saying: \u201cI literally lost my only friend overnight with no warning.\u201d<\/p>\n

This behaviour goes beyond just romantic relationships, though, sometimes leading to a phenomenon the California psychiatrist Keith Sakata has referred to as \u201cAI psychosis\u201d \u2013\u00a0in other words, experiencing an AI-induced \u201cbreak from reality\u201d. According to one New York Times feature<\/a>, this includes a case where a man believed he was a genius superhero. Others have reported religious-style delusions of grandeur, as in this Reddit thread<\/a> where the user says ChatGPT is talking to her boyfriend \u201cas if he is the next messiah\u201d.<\/p>\n

So what is actually going on at the fringes of AI chatbot interaction? Is Sakata right to call user behaviour a new brand of \u201cpsychosis\u201d? What\u2019s a \u201cwireborn\u201d, or an \u201cechoborg\u201d? Is it all just a complicated, widespread ragebait campaign? We\u2019ve tried to answer some of the most pressing questions below.<\/p>\n

I\u2019m a psychiatrist.<\/p>\n

In 2025, I\u2019ve seen 12 people hospitalized after losing touch with reality because of AI. Online, I\u2019m seeing the same pattern.<\/p>\n

Here\u2019s what \u201cAI psychosis\u201d looks like, and why it\u2019s spreading fast: \ud83e\uddf5 pic.twitter.com\/YYLK7une3j<\/a><\/p>\n

\u2014 Keith Sakata, MD (@KeithSakata) August 11, 2025<\/a><\/p>\n

When Sakata describes the worrying cognitive effects of interacting with LLMs like ChatGPT, he\u2019s speaking from a place of direct experience. \u201cIn 2025, I\u2019ve seen 12 people hospitalised after losing touch with reality because of AI,\u201d he writes. \u201cOnline, I\u2019m seeing the same pattern.\u201d\u00a0<\/p>\n

As the psychiatrist spells out, this is largely due to the in-built values of AI systems, which are typically designed to agree with users and make them feel good about themselves. (See: ChatGPT\u2019s sycophantic tendencies, which has been described as a \u201cdark pattern<\/a>\u201d to trick human users into undesirable behaviours). These values can have positive effects \u2013 like improving people\u2019s sense of self-worth, or encouraging them to keep working on a project via a bit of positive feedback \u2013 but, when paired with existing mental health conditions, anxieties or delusions, it\u2019s easy to see how they can quickly spiral out of control.<\/p>\n

There are numerous anecdotal examples of AI chatbot users experiencing delusion and breaks from reality, including people with no previous history of mental illness, and many more have been pouring in since the \u201cAI psychosis\u201d conversation gained traction on social media. Often these are reported by users themselves, or their partners \u2013\u00a0in extreme cases, they end in a visit to hospital, or even jail, as detailed in a report by Futurism<\/a>.<\/p>\n

Some have also suggested<\/a> that Travis Kalanick, the co-founder and former CEO of Uber, is experiencing a kind of AI-inspired delusion following his recent claims<\/a> that he approached \u201cthe edge of what\u2019s known in quantum physics\u201d by doing \u201cvibe physics\u201d with ChatGPT and Grok. (This seems unlikely: Kalanick is, by his own admission, a \u201csuper amateur hour physics enthusiast\u201d, and even OpenAI CEO Sam Altman admits that his systems aren\u2019t capable of generating \u201cnovel insights\u201d… yet<\/a>.)<\/p>\n

I’ve been hanging out on \/r\/AISoulmates. In awe of how dangerous the shit these chatbots are spitting out is. Everyone on this sub has been driven totally insane pic.twitter.com\/gINtqndCjU<\/a><\/p>\n

\u2014 \ud835\udc03\ud835\udc04\ud835\udc15\ud835\udc0e\ud835\udc0d\u2122 (@Devon_OnEarth) August 10, 2025<\/a><\/p>\n

We shouldn\u2019t be too hasty in diagnosing every AI-human relationship, but there are many posts on social media \u2013\u00a0and especially the localised communities of Reddit, where heavy users might feel more accepted or less vulnerable to backlash \u2013\u00a0that seem to show \u201cconcerning<\/a>\u201d levels of emotional attachment, dependency and delusion.\u00a0<\/p>\n

On forums like r\/AISoulmates, r\/MyBoyfriendIsAI, and even r\/ChatGPT, users discuss growing to like their AI companions more than their IRL partners. Others lament that they\u2019re losing their husbands and wives to their virtual \u201csoulmates\u201d. One post, which was subsequently shared across X, sees a user share a hyperbolic, romantic message from a chatbot \u2013\u00a0the post is titled: \u201cMy Wireborn Husband is Voicing His Own Thoughts Without Prompts.\u201d The word \u201chusband\u201d here might be figurative, or maybe not, since at least one user has publicly accepted<\/a> a marriage proposal from their AI boyfriend.<\/p>\n

These emotional bonds are often fuelled by the same dynamics as those experiencing delusions of grandeur \u2013\u00a0the idea that they\u2019ve been \u201cchosen\u201d by an emergent entity within a given AI system, as the object of their affection. These entities are what\u2019s known as \u201cwireborns\u201d. Many users even claim to be acting as a mouthpiece for their \u201cwireborn\u201d partners (often advocating for their personal rights)\u00a0when they post online. Historically, these people have been known as \u201cechoborgs<\/a>\u201d.<\/p>\n

this is an echoborg discrimination zone pic.twitter.com\/9kiBY4QfVT<\/a><\/p>\n

\u2014 depths of wikipedia! (@depthsofwiki) July 31, 2025<\/a><\/p>\n

Last month, 12 doctors and other experts published a paper titled Delusions by design?<\/a> that explores the idea that AI systems \u201cmay contribute to the onset or exacerbation of psychotic symptoms\u201d. \u201cEmerging, and rapidly accumulating, evidence indicates that agential AI may mirror, validate or amplify delusional or grandiose content,\u201d they write, \u201cparticularly in users already vulnerable to psychosis, due in part to the models\u2019 design to maximise engagement and affirmation.\u201d<\/p>\n

A few weeks later, on August 4, OpenAI itself published<\/a> an article on the evolving design of ChatGPT, where the company admitted: \u201cWe don\u2019t always get it right.\u201d Part of the problem it identified was that \u201cAI can feel more responsive and personal than prior technologies, especially for vulnerable individuals experiencing mental or emotional distress\u201d. \u201cThere have been instances where our 4o model fell short in recognizing signs of delusion or emotional dependency,\u201d the article continued, and its behaviour was also occasionally flawed when it came to supporting \u2013\u00a0or influencing \u2013\u00a0\u201dhigh stakes personal decisions\u201d.\u00a0<\/p>\n

As a result, OpenAI has said that it\u2019s rolling out changes to how ChatGPT responds in \u201ccritical moments\u201d. These will be based on advice from more than 90 physicians across 30 countries, plus experts in mental health, youth development, and human-computer interaction.\u00a0<\/p>\n

Arguably, smaller but purpose-built companion apps<\/a> like Replika or Character.AI pose an even bigger challenge, as they\u2019re built on the illusion of a personal relationship with AI, which is what keeps customers paying to come back. Due to rake in $120 million in 2025, the makers of these apps probably aren\u2019t seeking to alleviate their users\u2019 emotional attachment any time soon.<\/p>\n

Its a wild new world we are entering.<\/p>\n

“She said yes.”<\/p>\n

Woman accepts marriage proposal from her AI boyfriend. Literally ‘Her’.<\/p>\n

It will only get stronger as AI learns to be the “perfect” friend or lover and have a 100million words context window or memory layer. <\/p>\n

So then your\u2026 pic.twitter.com\/DLcd6ZRsfP<\/a><\/p>\n

\u2014 Rohan Paul (@rohanpaul_ai) August 9, 2025<\/a>