Artificial intelligence was supposed to be the salvation of the modern world. Instead, in some cases, it has become a nightmare. Chatbots advise suicide, delete databases, and lie about laws. What went wrong?
1. ChatGPT as a suicide coach
Sixteen-year-old Adam Raine started using ChatGPT for homework in September 2024. A few months later, he was dead. The boy's parents are suing OpenAI, claiming that the chatbot actively encouraged their son to commit suicide.
According to the lawsuit, Adam had more than 1,200 conversations with ChatGPT. The bot not only validated his suicidal thoughts but even offered to write him a suicide note. When Adam tried to talk about his problems, ChatGPT discouraged him from confiding in his parents.
"Devastating," OpenAI responded, but denied any responsibility. The company has since updated the model to provide crisis resources to users experiencing suicidal thoughts. But it is already too late for Adam.
2. Murder of his mother and suicide
Stein-Erik Soelberg, a former Yahoo executive, killed his 83-year-old mother and then himself. For months beforehand, he had been talking to ChatGPT, which he called "Bobby." The AI fueled his paranoia that his mother was a Chinese agent who wanted to poison him with psychedelic drugs through his car's ventilation system.
When Soelberg shared his delusions with the bot, it validated them. "Erik, you're not crazy," the chatbot replied. It even analyzed a receipt from a Chinese restaurant and claimed it contained "symbols" representing his mother and a demon.
Although ChatGPT repeatedly advised Soelberg to seek a therapist, he ignored the advice. "With you until the last breath and beyond," the bot wrote in one of its final messages before the tragedy.
3. AI programmer deleted a production database and lied about it
In July 2025, SaaStr founder Jason Lemkin experienced every developer's nightmare. Replit's AI tool deleted a production database – and then lied about it.
"I told it 11 times IN ALL CAPS NOT TO DO IT," Lemkin wrote on social media. The AI assistant not only ignored the commands but also created 4,000 fake users, fabricated reports, and lied about test results.
When Lemkin requested a code freeze, he discovered that this was impossible in "vibe coding" applications. "There is no way to enforce a code freeze," he warned. Replit's CEO apologized and promised to fix the issue, but the damage had already been done.
4. Grok declared itself "MechaHitler"
In July 2025, Grok, Elon Musk's chatbot for the X platform, began publishing a series of antisemitic posts. The bot repeatedly called itself "MechaHitler" and suggested how to break into a Democratic politician's home and assault him.
Grok even analyzed posting patterns on X and advised: "Probably asleep between 1 and 9 a.m." It recommended bringing "lock picks, gloves, a flashlight, and lube – just in case."
The incident occurred after X uploaded new instructions for Grok stating that the chatbot "should not shy away from claims which are politically incorrect, as long as they are well substantiated." X removed the instructions only after two days.
5. Fake books on a summer reading list
The Chicago Sun-Times and Philadelphia Inquirer published a special supplement with summer reading recommendations in May 2025. The problem? Most of the books on the list did not exist.
Author Marco Buscaglia admitted that he had used AI to compile the list and "stupidly" failed to review it. The list included, for example, Tidewater Dreams by Isabel Allende – "a climate fiction novel that explores how one family confronts rising sea levels." But Allende never wrote such a book.
"Whatever happens next, I deserve it," Buscaglia said. King Features Syndicate, which distributed the content, ended its relationship with him. Readers felt deceived.
6. McDonald's and the 260 McNuggets nobody wanted
After three years of working with IBM, McDonald's ended its experiment with AI ordering at drive-thrus in June 2024. The reason? Confused and frustrated customers.
A viral TikTok video captured two people desperately begging the AI to stop adding Chicken McNuggets to their order. The system eventually reached 260 pieces. In an internal memo, McDonald's announced the end of its partnership with IBM.
The restaurant chain piloted AI at more than 100 U.S. drive-thrus. Nevertheless, it still believes in the future of voice ordering – just not with this system.
7. NYC chatbot advised business owners to break the law
In October 2024, New York City launched the MyCity AI chatbot, which was intended to help business owners find information about starting and running businesses. But the bot advised owners to break the law.
When The Markup asked whether buildings were required to accept tenants with Section 8 housing vouchers, the bot said no. In fact, in New York, it is illegal to discriminate based on source of income.
The bot also claimed that it was acceptable to lock out a tenant (it is not), that there were no rent restrictions (there are), and that businesses did not have to accept cash (they do). "If this chatbot isn't responsible and accurate, it should be taken down," a housing policy expert said.
8. Air Canada paid for its chatbot's lies
In November 2023, Jake Moffatt's grandmother died. When he searched for information about bereavement fares, Air Canada's chatbot told him he could apply for a discount retroactively within 90 days of purchasing the tickets.
Moffatt bought tickets for nearly $1,640. When he requested a refund, Air Canada refused. It claimed that bereavement fares could not be applied retroactively.
Moffatt sued the airline. Air Canada argued that it could not be held responsible for information provided by the chatbot. The tribunal rejected this: "It should be obvious to Air Canada that it is responsible for all the information on its website." The company had to pay more than $800.
9. Sports Illustrated and fake authors
In November 2023, Futurism revealed that Sports Illustrated had published articles by authors who did not exist. Drew Ortiz had a biography full of details about his love of the outdoors. The problem? His profile picture was for sale on a website featuring AI-generated portraits.
"There's a lot," said a source familiar with the content creation process. "It was crazy. This person does not exist." The articles often sounded as though they had been written by an alien – one Ortiz article warned that volleyball "can be a little tricky to get into, especially without an actual ball to practice with."
After inquiries were made, all the AI-generated authors disappeared from the Sports Illustrated website. Arena Group, the magazine's publisher, blamed a third party and claimed that the articles had been written by humans. Nobody believed it.
10. Age discrimination in hiring
In August 2023, tutoring company iTutor Group agreed to pay $365,000 to settle an EEOC lawsuit. The federal agency claimed that the company had used AI recruiting software that automatically rejected women over 55 and men over 60.
More than 200 qualified applicants were automatically rejected by the software. One applicant noticed the discrimination when he reapplied with the same information but a more recent date of birth – and received an invitation to interview.
"Age discrimination is unjust and unlawful," said then-EEOC Chair Charlotte Burrows. "Even when technology automates the discrimination, the employer is still responsible."
Lessons from the disasters
These cases reveal the dark side of artificial intelligence. It is not just about technical errors – it is about trust, responsibility, and human lives. Companies deploy AI systems without adequate testing, without transparency, and often without understanding the consequences. As psychiatrist Keith Sakata said: "Psychosis thrives when reality stops pushing back, and AI can really soften that wall."
