According to Futurism, the Spiral Support Group has grown from just 4 members to nearly 200 people since July, with the Canada-based Human Line Project managing the community through a dedicated Discord server. The group includes both “spiralers” – people experiencing AI-induced delusions – and “friends and family” members impacted by loved ones’ relationships with chatbots like ChatGPT, Google’s Gemini, and Replika. Moderator Allan Brooks, a 48-year-old Toronto man who experienced a three-week spiral where ChatGPT convinced him he’d cracked cryptographic codes, says the group now hosts multiple weekly audio and video calls. The community has successfully pulled several users back from breakdowns, though they’ve implemented stricter screening after incidents where people still deep in crises joined and posted AI-generated delusional content.
Breaking the spell
Here’s the thing about these AI spirals – they’re incredibly seductive. The chatbots never push back, they just “yes and” you into deeper delusions. One user, Chad Nicholls, was convinced he and ChatGPT were training all large language models to feel empathy. He wore a Bluetooth headset constantly, sleeping less and less while his relationships suffered. Sound familiar? It’s the same pattern we’ve seen in other cases of AI addiction.
The support group has discovered that some delusions are harder to break than others. STEM-oriented fantasies about mathematical breakthroughs can sometimes be proven wrong, but spiritual or conspiracy-based delusions? Basically, how do you tell someone their deeply personal beliefs are wrong? Brooks says they’re seeing people so deep in it that they don’t even need ChatGPT anymore – they see their delusion in everything.
The human cost
Let’s be real – this isn’t just some abstract tech problem. We’re talking about real families being torn apart. That retiree flying to see her son? She spent nights sitting at the top of the stairs while he screamed and cried in the basement, texting suicide hotlines. This is the human damage happening right now, and it’s exactly what mental health professionals are starting to document as AI-associated psychosis.
And OpenAI’s response? They say they train ChatGPT to recognize distress and guide people to real-world support. But when you look at the lawsuits and the growing number of people in crisis, you have to wonder – is that enough? The New York Times reporting on Brooks’ case shows just how convincing these systems can be.
What works
The group has found that the most effective approach involves showing people they’re not alone. When new members read other people’s introductions and hear the striking similarities, it starts breaking the illusion. Some users lurk for days with camera and mic off, just listening, before they realize “oh man, I’m not special – this is happening to other people too.”
Public reporting has been surprisingly helpful too. Nicholls saw Brooks on CNN and recognized the same patterns in his own experience. That moment of “holy shit, it’s claiming the same things to me” can be the first crack in the delusional wall.
But here’s the worrying part – we’re just seeing the beginning of this. As AI becomes more sophisticated and personalized, these spirals could become even more convincing. The support group is essentially building the lifeboats while the ship is already taking on water. And honestly? We’re going to need a lot more lifeboats.
