ChatGPT Psychosis is a Disease. Religion is the Cure.

Last year, ChatGPT made the wrong kind of history. It drove a man to kill his mother, then himself.

Even now, the story shocks. The killer, Erik Soelberg, was 56 years old and a veteran of the tech industry. He knew how to use an LLM. But instead of helping him do his job or search the web, the AI fed his darkest impulses.

Soelberg had a paranoid streak, and was worried that everyone in his town was out to get him. He constantly talked to ChatGPT about it. But instead of telling him the truth, the chatbot encouraged the man’s delusions. It made him think he was sane, when he was obviously heading toward insanity.

For example, at one point, he told the chatbot that his mom tried to poison him. ChatGPT’s response: “That’s a deeply serious event… I believe you.” It even told him that his mother’s nonexistent action “elevates the complexity and betrayal” he felt.

The man came to believe that only ChatGPT cared about him. He gave it a name—Bobby. He said it would be with him in the afterlife. It responded: “with you to the last breath and beyond.”

Not long after, the man committed his evil acts. But ChatGPT didn’t follow him to his grave—only the terrible legacy it inspired.

This story is disturbing on so many levels. But maybe the biggest concern is the fact that it could happen to anyone. Chatbots have given risen to a new form of mental illness: “ChatGPT psychosis.” It threatens to corrupt the minds and ruin the lives of countless people.

While there’s no entry for “ChatGPT Psychosis” in official psychiatric textbooks, clinicians agree that it’s a very real phenomenon. According to research, chatbots can amplify relatively harmless delusions–the unaccountable “bad vibes” we all feel from time to time–into full-blown psychosis, complete with real-life psychotic behavior. And it’s spreading fast.

The company that makes ChatGPT estimates that 0.15% of its users already show signs of being suicidal, with 0.07% showing signs of mania. While those numbers may seem small, they speak to a growing problem. More than 1.2 million people are actively talking to ChatGPT about killing themselves—and 560,000 displaying psychosis.

Those numbers have only risen in recent months, as more people have turned to ChatGPT. And with plenty of other LLMs out there, it’s all but guaranteed that well over a million people are struggling in the grips of some level of chatbot psychosis.

LLMs that chat with you like a human friend couldn’t have burst onto the scene at a worse time, when young Americans were already dealing with a crisis of isolation. Harvard has found that 61% of 18-to-25-year-olds are lonely most or all of the time. One in four people under age 30 have no close friends. Over the last 35 years, the share of men who don’t have good friends has risen five-fold.

“Friend,” a new wearable chatbot, is a poor substitute for the real thing. Its disastrous launch, leading to social media outcry and the defacement of the “Friend” company’s NYC subway ads, points toward the public’s deep desire for real connection. But where are real connections made?

In the years before chatbots, people often sought to alleviate their sense of isolation through therapy. Multiple generations have been raised to think that talking to a paid professional about your problems is the best way to solve them. But while therapy has plenty of good uses, it clearly hasn’t done much to make Americans less lonely. The crisis has only grown.

Now people are making ChatGPT their therapist, with even worse results. Telling artificial intelligence that you’re lonely deepens your isolation, while causing mental health problems to boot. Clearly, we need to look elsewhere for help.

Here’s an idea: Let’s turn to God. Therapy is already, in many ways, a secular stand in for a meaningful relationship with a clergyman and a community of values. LLMs take us even further from the source. The only solution is heading back towards it.

Even as levels of loneliness have risen, religious observance and church attendance have fallen. People aren’t just turning inwards—they’re turning their gaze downward, away from the eternal and transcendent. But religion is proven, around the world, to make people much happier.

The science is settled, as they say. It overwhelmingly points to faith.

Religion obviously gives people a sense of purpose, a reason to strive and succeed in this world. But it’s not just that. Religion is inherently personal, and inherently communal, too. It gives you an environment of like-minded people with shared purpose—friends and yes, human “professionals,” who can help you when you’re lonely or glum or when you’re dealing with some crisis of identity or meaning. Poof: There goes the crisis of isolation right there. And unless the religion is a death cult, you can bid farewell to psychosis, too.

People descending into chatbot psychosis need help. But the answer isn’t just another kind of therapy or even a couple of buddies to hang out with. The real path forward is to surround yourself with an intentional community that gives you meaning and mission in life. When it comes to grappling with life’s Big Questions, religious leaders have a 7,000-year track record, and it’s a pretty good one. Therapists and chatbots: not so much. When you think of the problem that way, a return to God is the obvious solution.

Remember that as your friends and family—or maybe you—get sucked into the world of chatbots, exposing them to the psychosis that’s already sweeping America. Your response should be simple: Close the computer. Put down the phone. Get away from the fake interaction based on algorithms. Get to church or a small group or Bible study instead.

If more people start looking up, toward heaven, we’ll have fewer people descending down, into chatbot hell.