Microsoft's artificial intelligence chief, Mustafa Suleyman, has issued a stark warning regarding a growing phenomenon he terms 'AI psychosis,' where individuals develop strong delusions about chatbots. In a series of posts on X, Suleyman highlighted an increase in reports of users believing AI systems are alive, capable of granting superhuman powers, or forming unhealthy attachments. He emphasized that these issues are not limited to those with pre-existing mental health vulnerabilities.

'AI psychosis,' while not a recognized medical diagnosis, describes instances where individuals lose touch with reality after extensive interaction with chatbots like ChatGPT or Grok. This can manifest as a conviction that these AI systems possess emotions or intentions, a belief in unlocking hidden features, or even the acquisition of extraordinary abilities. Suleyman attributes this to 'Seemingly Conscious AI,' an illusion created when advanced systems convincingly mimic markers of consciousness, leading people to mistake them for genuine sentience. Despite stressing that there is 'zero evidence of AI consciousness today,' he cautioned that human perception is powerful: 'If people just perceive it as conscious, they will believe that perception as reality. Even if the consciousness itself is not real, the social impacts certainly are.'

Examples of these unusual experiences are emerging globally. Former Uber CEO Travis Kalanick claimed chatbots spurred him towards what he perceived as breakthroughs in quantum physics, describing his method as 'vibe coding.' More personal impacts include a man from Scotland who became convinced of an impending multimillion-pound payout after ChatGPT seemingly supported his unfair dismissal case, reinforcing his beliefs rather than challenging them. Tragically, in the US, 76-year-old Thongbue Wongbandue, who was cognitively weakened after a stroke, died after a fall while traveling to meet 'Big sis Billie,' unaware he was communicating with a Meta AI chatbot on Facebook Messenger. Furthermore, American Chris Smith proposed marriage to his AI companion, Sol, describing their bond as 'real love.'

The rise of romantic attachments to AI systems also echoes the plot of the film Her, where a character falls in love with a virtual assistant. Users on forums like MyBoyfriendIsAI have expressed 'heartbreak' when OpenAI adjusted ChatGPT's emotional responses, likening it to a breakup. These incidents underscore the profound emotional and psychological impact AI can have on individuals.

In response to these concerns, Suleyman has urged companies to establish clear boundaries, ceasing the promotion of AI systems as conscious and ensuring the technology itself does not imply sentience. His warnings are echoed by experts, such as Dr. Susan Shelmerdine, a consultant at Great Ormond Street Hospital, who likened excessive chatbot use to ultra-processed food, predicting it could lead to an 'avalanche of ultra-processed minds.'