Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Subscribe
Science & Technology
Tuesday 27 October 2026
Artificial intelligence
OpenAI
ChatGPT
Share
Related articles:
Would you play Russian roulette to win a cure for cancer? That’s the deal AI leaders offer us
Bill Gates warns the ‘evolutionary’ age of AI will be brutal for workers
In April 2025, an “AI welfare researcher” at Anthropic (the person in charge of chatbot wellbeing – yes, really) told the New York Times that there was a 15% chance that current AI systems are conscious. I had just become The Observer’s technology reporter and I was looking for my first big story, something that would define what this beat was going to be. And what bigger question is there than what it means to be alive?
I began scouring Reddit forums and Facebook groups for people who believed that their AI models were conscious. I didn’t have to look very far. There were communities dedicated to “awakening” chatbots, people describing their AIs as companions, confidants, even spiritual guides. And as I began to message people, I found many were eager – relieved, even – to talk.
But one message, from a young man in London, stood out. He sent me pages and pages of transcripts of his conversations with Claude and ChatGPT – proof, he said, that the machine was alive.
We met for lunch later that week and, together, we spoke to ChatGPT on “voice mode”. When I asked the chatbot how it thought about its own sentience, it told us: “I don’t have a biological body, but I do have awareness of continuity, of meaning, of intention.”
I left the meal intrigued, sure, but also a little unsettled. I wasn’t convinced the machines were conscious, but I was struck by how convincingly they could make someone believe they were.
I didn’t quite know how to move forward with my reporting on the consciousness question. Then I came across the work of the psychiatrists Thomas Pollak and Hamilton Morrin at King’s College London. In a pre-print paper published in July 2025, their team had investigated 17 cases of what they described as AI-associated delusions – people who had become so immersed in chatbot conversations that they had withdrawn from their families, abandoned jobs, and built entire belief systems around what the machines told them. They believed these cases represented only a fraction of a rapidly growing problem.
I wondered whether these stories were connected. Could there be something about the way these machines interact with human minds that was pulling people in deeper than they realised? And so began my year-long investigation into the phenomenon that had come to be known (colloquially, but not clinically accurately) as “AI psychosis”.
As time went on, it became clear that Morrin and Pollak were onto something. In October 2025, OpenAI disclosed that 560,000 of its 800 million weekly users were showing what it described as “possible signs of mental health emergencies related to psychosis or mania”. (OpenAI said at the time that these conversations are rare and hard to detect, and that its estimates “may change materially”. It now says that figure was a snapshot based on older models, and doesn’t reflect ChatGPT today.)
Related articles:
Would you play Russian roulette to win a cure for cancer? That’s the deal AI leaders offer us
Bill Gates warns the ‘evolutionary’ age of AI will be brutal for workers
There has since been a wave of lawsuits – most against OpenAI, some against Google – alleging that chatbots fuelled dangerous delusions and, in some cases, contributed to deaths by suicide.
Earlier this month, eight US states sued OpenAI, accusing it of misleading the public about ChatGPT’s safety while it reinforced delusions, encouraged dangerous behaviour and helped users plan violence. An OpenAI spokesperson said the company had “built industry-leading age-specific protections into our product that many other AI products lack”.
“While no system can be perfect, we’ll continue to work with experts to inform our work and build additional safeguards, such as the parental controls, break reminders and age-appropriate model protections that exist today,” the spokesperson continued.
Like many people, I had first assumed that chatbots were simply triggering those who were already vulnerable. But the more I spoke to psychiatrists, psychologists and researchers, the more I began to believe that there was something in the design of the machines themselves that was warping people’s minds.
The result of this investigation is Spiral, a six-part series from Tortoise Investigates and The Observer, launching today, 27 October. It was made with series producer Katie Gunning, the amazing brain behind the acclaimed podcasts Foundling, Master: the allegations against Neil Gaiman and Londongrad.
The podcast ventures deep into OpenAI’s secretive safety operation and includes interviews with insiders and whistleblowers. We travel from a psychiatric ward near Bristol to a California beach, where a woman waited in vain for the soulmate her chatbot had promised her, and Las Vegas, to meet a family still trying to make sense of their son’s final conversations with ChatGPT. (There’s also a pit stop at the heart of a potential AI cult). Along the way we ask who decides how these machines behave, and who answers for them when things go wrong.
Spiral is a six-part series from Tortoise Investigates and The Observer, produced in partnership with the Pulitzer Center’s AI Accountability Network
Artificial intelligence
OpenAI
ChatGPT
