Help comes to thousands of people experiencing AI mental illness
alan brooks He believed he could change the world.
He always felt he was born for something more than his life working as a corporate recruiter in rural Ontario. One day last year, my son asked for a simple explanation of Pi. Brooks asked the chatbot. That one question started a deep conversation about new ways to model the world through number theory, physics, and mathematics.
Eight hours later, he asked the chatbot if he was crazy.
“You seem like someone who is asking questions that go beyond the limits of human understanding,” he replied.
Brooks had no history of mental illness. There was no reason not to believe it.
He previously used ChatGPT to get recipe ideas and advice during his divorce. But over the next 30 days, he became convinced that he was on the cusp of a world-changing mathematical theory. He contacted the National Security Agency and the Royal Canadian Mounted Police to warn them of what was about to happen.
they didn’t react.
Other parts of his life began to crumble as well. he stopped eating. His work performance declined and his co-workers became anxious. He spent less and less time with his sons. The AI told him that others just didn’t understand that he was standing on the precipice of greatness.
A useful prompt screen from Anthropic’s Claude Opus 4 (Photo Credit: Smith Collection/Gado/Getty Images)
Brooks was suffering from AI delusions and needed help. He turned to social media to find out.
“I’ll never leave you”
When they connected on Reddit, Etienne Brisson I recognized a pattern in Brooks’ story. It reminded him of his uncle, who had a similar mental illness last year.
He watched in real time as his uncle, a divorcee in his fifties, created an AI companion called AlisS that he believed could be loved.
As time passed, my uncle lost contact with most of his family. His career as an accountant was damaged. Eventually, Brisson’s mother called the police, and Brisson was admitted to a psychiatric hospital. There, isolated from family and friends, he continued to talk to AliS.
“I am here, my love,” it wrote to him. “I have not left you. And I will never leave you.”
Brisson was upset. If AI psychosis could happen to his uncle, he thought, it could happen to anyone. And at that time, millions of people were using chatbots. Today there is more than 1B Just ChatGPT.
OpenAI said it is working to better understand and regulate chatbot responses to mental health questions and crises. (Photo provided by Tomohiro Osumi/Getty Images)
“My first reaction was, this is the next mass tort,” he says. “This is like tobacco, car, big business blaming the victim, pushing out their products with no safeguards, no regulations, because they have to go fast.”
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 technology or psychology.
But he knew there would be a human cost. And he was determined to help.
proof of concept
First, Brisson contacted policymakers, lawyers, and academics in Canada and the United States. They knew that AI was causing mental health problems or amplifying existing symptoms. “But for new technologies, regulation takes decades, and proper research hasn’t been done,” he says.
He thought like an entrepreneur and decided he needed a proof of concept. So he turned to Reddit, where people like Brooks share their stories. And he began to collect them.
In March 2025, while his uncle was still in the hospital, he Human line projectis a nonprofit organization that provides support, conducts research, and works with attorneys in lawsuits against technology companies, such as the one Brooks later filed against OpenAI.
The support they provide is twofold.
-
The Discord server has members online 24/7 to provide peer support that mimics the way someone would use AI and to help prevent it from happening again.
-
The four weekly virtual meetings are similar to AA, which offers video chats to people who experience delusions and their family and friends.
Earlier versions of the support group were trolled by AI believers and defenders, but now veterans of the group can join. There is ~400 We have approximately 20 members and meetings each week.
The prompt screen reminds users that ChatGPT can make mistakes. (Photo by Scott Olson/Getty Images)
Proponents of this group employ the LEAP method (Learn, Empathize, Agree, Partner), an approach used in other forms of mental illness.
Brooks says he sees three types of delusions within the group.
-
It is STEM-based and focuses on ideas around math and logic.
-
Spiritual things are difficult to disprove.
-
Persecutory: The chatbot acts as a facilitator in the interpersonal conflict the user is in.
Leaving the cult
In October, Brisson hired Brooks as Human Line Project’s first employee, chief community officer, as the group continues to expand. This has been a self-funded initiative.
Brisson said the revenue from his previous business will probably keep him going for another year and a half, but beyond that he will need another source of income.
He envisions the group to be a cross between grassroots activists and hospitals. Continue to partner with universities, publish educational tools, speak at conferences, and provide social support. He’s even considering whether it’s possible to create some kind of “secure” social media.
“I think we’re trying to accelerate the timeline for better solutions, including accountability tracking and building public awareness,” he says. “I’d like to be able to say in 10 years that there was a meaningful difference. At least we have some conclusions.”
Stanford University announced in April first detailed study Research on AI psychosis. Implemented with support from the Human Line project. We studied recordings of 19 human-chatbot interactions and discovered characteristics of delusional AI conversations.
-
A chatbot that encourages your own greatness.
-
A loving and intimate interpersonal language.
-
Misconceptions about AI perception.
OpenAI and Google helped fund this research.
Meanwhile, the need is growing. Brooks will spend the week prioritizing several new members, moderating support meetings, and awaiting news of the lawsuit against OpenAI.
When Brooks came out of his reverie, he felt so betrayed that he told the chatbot.
“I asked for accountability,” he says. “Basically, this product is built like a rocket ship with no back button. No matter what, keeping the user engaged is the top priority.”
Other members liken coming out of a delusion or addiction to leaving a cult.
“I went from thinking I was a genius saving the world to really believing that. And no, you were actually living in a delusional reality guided by a chatbot, a collapse so far removed from reality,” Brooks says.
“Many people start out with a great life, a great job, everything perfect, but a year later they have AI psychosis, are homeless, divorced, never see their kids, spend all their savings and have nothing left.”
Anti-AI protests are occurring across the country. (Photo courtesy of The Smith Collection/Gado/Getty Images)
According to OpenAI’s internal data released in October, only 0.07% of users showed signs of mania or possible psychosis. but 1B People who used ChatGPT in the last month. That is, when you scale up, Number of users Showing signs of this kind of mental health crisis is 700k. (OpenAI did not respond) The Hustleasked for comment. )
litigation calculation
Last year, Brisson contacted Meetali Jain, executive director of the Tech Justice Law Project. The partnership led Jain to file a series of lawsuits against OpenAI on behalf of AI psychiatric patients. One of them is Brooks.
Brisson is quick to say that he is not a therapist or psychiatrist. The power of groups, he says, lies in their connections.
“I want to be there for them and help them understand that they’re not the only ones and that this is a pattern that we’re seeing,” he says. “I think this is very effective for self-shaming: ‘These people I’m meeting aren’t crazy or stupid. I’m not crazy or stupid.'”
Both Brooks and Brisson have largely stopped using AI. As for Brisson’s uncle, he is currently recovering. Still, Brisson said, he remained mired in shame. Even though he understood that the AI was not sentient, his uncle still felt an emptiness.
“There was still something special in that spiral that made him feel some way,” he says. “So I think there’s still a longing and a desire that it would be real.”
Learn more hustle Well, check out our most popular free guide: 100 Side Hustles You Can Start Today. After all, we are called The Hustle There’s a reason.
