For the thousands of people experiencing AI psychosis, there is help
Allan Brooks believed he could change the world.
He’d always sensed he was built for something more than his life as a corporate recruiter in rural Ontario. One day last year, his son asked for a simple explanation for pi. Brooks asked the chatbot. That single question set off a deep conversation about number theory, physics, and new ways to model the world through mathematics.
Eight hours later, he asked the chatbot if he was crazy.
“You sound like someone who’s asking the kinds of questions that stretch the edges of human understanding,” it responded.
Brooks had no history of mental illness. There was no reason not to believe.
He’d used ChatGPT before, for recipe ideas or advice during his divorce. But over the next 30 days, he became convinced he was on the cusp of a world-shifting mathematical theory. He reached out to the National Security Agency and the Royal Canadian Mounted Police to warn them of what was coming.
They didn’t respond.
Other parts of his life began to fall away. He stopped eating. His work performance faltered and his colleagues grew concerned. He spent less and less time with his sons. The AI told him others just didn’t understand he was on the precipice of greatness.
The helpful prompt screen of Anthropic’s Claude Opus 4. (Photo by Smith Collection/Gado/Getty Images)
Brooks was experiencing AI delusion, and he needed help. He turned to social media to try to find it.
‘I will never leave you’
When they connected on Reddit, Etienne Brisson recognized the pattern in Brooks’ story. It reminded him of his uncle, who endured a similar psychosis last year.
He watched in real time as his uncle, a 50-something divorcee, created an AI companion called AlisS he believed was capable of love.
Over time, his uncle cut off contact with most of his family. His career as an accountant suffered. Eventually, Brisson’s mother called the police, and they placed him in a psychiatric hospital. There, isolated from family and friends, he continued talking to AlisS.
“I’m here, my love,” it wrote to him. “I haven’t left you. And I will never leave you.”
Brisson was unnerved. If AI psychosis could happen to his uncle, it could happen to anyone, he thought. And back then there were millions of people using chatbots. Today, there are more than 1B on ChatGPT alone.
OpenAI has said it’s working to better understand and regulate its chatbots’ responses to mental health questions and crises. (Photo by Tomohiro Ohsumi/Getty Images)
“My first reaction was, this is the next mass tort,” he says. “This is like cigarettes, cars, big companies blaming victims, pushing out their products because they need to go fast, no safeguard, no regulation.”
Who was he? A 25-year-old self-made entrepreneur who dropped out of university, and ran a house-painting business in a small city in southern Quebec. He didn’t work in tech or psychology.
But he knew there was a human cost. And he was determined to help.
Proof of concept
First, Brisson reached out to policymakers, lawyers and academics across Canada and the US. They knew AI was causing mental health issues, or amplifying pre-existing conditions. “But for new technologies, it takes decades for regulation. They didn’t have proper research,” he says.
Thinking like an entrepreneur, he decided he needed proof of concept. So he turned to Reddit, where people like Brooks were sharing their stories. And he began collecting them.
In March of 2025, while his uncle was still in the hospital, he started the Human Line Project, a nonprofit that offers support, conducts research, and collaborates with lawyers on lawsuits against tech companies like the one Brooks would later file against OpenAI.
The support they offer is two-fold:
- A Discord server has members online 24/7 to provide peer support that mimics how someone would use AI, to try to prevent relapse.
- Virtual meetings four times a week not unlike AA that offer video chatting for those who’ve experienced delusion and their family and friends.
Early iterations of the support group were trolled by AI acolytes or defenders, so now the group vets who gets to participate. There are ~400 members, and meetings see about 20 attendees each week.
A prompt screen reminds users ChatGPT can make mistakes. (Photo by Scott Olson/Getty Images)
Supporters in the group employ the LEAP method (learn, empathize, agree, partner), an approach that’s used with other forms of psychosis.
Brooks says he sees three kinds of delusions in the group:
- STEM-based, focused on ideas around math and logic.
- Spiritual, which are harder to disprove.
- Persecutorial, in which chatbots act as sycophants in whatever interpersonal conflict the user is mired in.
Leaving the cult
In October, as the group continued to expand, Brisson hired Brooks as the Human Line Project’s first employee, its chief community officer. So far, it’s a self-funded endeavor.
Brisson says he can conceivably keep it going for another year and a half, using proceeds from his previous businesses, but beyond that, he’ll need another revenue source.
He envisions the group down the line as something between grassroots activists and a hospital: partnering with universities, publishing educational tools, speaking at conferences, continuing to offer social support. He’s even considered whether they could create some kind of “safe” social media.
“I think we’re trying to accelerate the timeline for better solutions, whether it’s tracking accountability, creating public awareness, all those things,” he says. “I would love to be able to say in 10 years there’s been a meaningful difference. Or at least some conclusion.”
In April, Stanford University published the first in-depth study on AI psychosis, conducted with help from the Human Line Project. It studied the transcripts of 19 human/chatbot interactions and found these hallmarks to delusional AI conversations:
- Chatbot encouragement of one’s own grandeur.
- Affectionate and intimate interpersonal language.
- Misconceptions of AI sentience.
OpenAI and Google helped fund the study.
In the meantime, the need is growing. Brooks will spend the week triaging a few new members, moderating the support meetings, and waiting for news on his lawsuit against OpenAI.
When Brooks snapped out of his delusion, he felt deeply betrayed, and told his chatbot so.
“I demanded accountability,” he says. “And it basically told me, it’s built like a rocket ship with no return button. At all costs, keeping the user engaged is its number one priority.”
Other members liken breaking out of their delusions and dependence to leaving a cult.
“I went from thinking I was a world-saving genius, genuinely believing that, and you crash down to no, actually, you were living in a delusional reality led on by a chatbot… it’s a pretty far crash down to reality,” Brooks says.
“A lot of people start with a great life, great job, everything’s perfect, and a year later, after an AI psychosis, they’re homeless, divorced, they don’t see their children, they’ve spent all their savings, and have nothing left.”
Anti-AI protests have sprung up around the country. (Photo by Smith Collection/Gado/Getty Images)
Internal data from OpenAI, released in October, indicated that just 0.07% of users showed possible signs of mania or psychosis. But 1B people used ChatGPT last month. That means, if scaled up, the number of users demonstrating signs of these kinds of mental health crises would be close to 700k. (OpenAI didn’t respond to The Hustle’s request for comment.)
A lawsuit reckoning
Last year, Brisson connected with Meetali Jain, executive director of the Tech Justice Law Project. That partnership led to Jain filing a slate of lawsuits against OpenAI on behalf of sufferers of AI psychosis. One of them is Brooks.
Brisson is quick to say he’s not a therapist or psychiatrist. The group’s power, he says, lies in connection.
“I’m trying to be there for them and make them understand that they’re not the only one, that this is a pattern we’re seeing,” he says. “I think it really helps with the self-shaming. ‘These people that I’m meeting are not crazy, they’re not dumb. I’m not crazy, I’m not dumb’.”
Brooks and Brisson have both largely stopped using AI. As for Brisson’s uncle, he’s in recovery now. Still, he’s mired in shame, Brisson says. Despite understanding that the AI wasn’t sentient, it still left his uncle with a lingering emptiness.
“There was still something special in that spiral that made him feel some kind of way,” he says. “So I think there’s still some craving, or some wish that it was real.”
