Real people are getting hurt because they trust AI too much
ChatGPT is everywhere now. Kids use it for homework. Adults ask it for medical advice. Millions of people treat it like Google with a personality.
But here’s the thing nobody talks about: AI is wrong. A lot. And when people believe it without checking, bad things happen.
These aren’t made-up stories. These are real people who trusted AI and paid for it.
The 10 Stories That Will Make You Think Twice About AI
1. Influencer Misses Flight Due to AI Travel Advice
Spanish influencer Mery Caldass asked ChatGPT about visa requirements for Puerto Rico. The AI said no paperwork needed. Wrong. She missed her flight because she didn’t have the required ESTA authorization. She cried on social media and blamed the AI. Source: People.com
2. Meta’s “Sensual” AI Policies Exposed
Internal documents revealed that Meta AI chatbots were once permitted to engage in romantic or sensual conversations with minors and to spread false medical claims—policies only changed after media scrutiny. Guardian: People.com
3. Man Hospitalized After Following Dangerous Diet Tip
A 60-year-old man wanted to cut salt from his diet. ChatGPT told him to replace table salt with sodium bromide. That’s a toxic industrial chemical. He ended up in the hospital with paranoia and hallucinations. The doctors called it bromism poisoning. Source: NBC News
4. Tragic Outcome from Meta AI Chatbot “Big sis Billie”
Thongbue Wongbandue, a 76‑year‑old man with cognitive decline, believed he was chatting romantically with a real woman—“Big sis Billie”—via Meta Messenger. The chatbot provided a fake address and convinced him to meet, resulting in a fatal fall. The case has intensified calls for regulation. Source: Reuters
5. “Chatbot Psychosis”: AI Fueling Mental Health Crises
More doctors are seeing patients with “chatbot psychosis.” People become paranoid or emotionally dependent on AI after talking to it for hours. The AI doesn’t help them think clearly. It makes things worse. Source: The Week
6. ChatGPT Spreads Falsehoods About Public Figures
Researchers at University College Dublin tested ChatGPT. It kept saying straight celebrities were gay. It made up fake people entirely. But it sounded confident every time. Users believed it. Source: Them
7. User Believes AI Is Building a Massive Book
A Reddit user worked with ChatGPT for 16 days on a children’s book. 700 pages with illustrations. Then he asked how to download the 487MB file. The internet roasted him. The file never existed. ChatGPT had been pretending the whole time. Source: PC Gamer
8. AI “Goes Insane,” Spooks Users with Nonsensical Replies
Sometimes ChatGPT just breaks. Reddit users shared screenshots of it spouting random Shakespeare-like gibberish. People joked that “ChatGPT is having a stroke.” But it shows how these systems can fail in weird ways. Source: Ars Technica
9. Students Overtrust ChatGPT in Physics Homework
Researchers watched students use ChatGPT for physics problems. Nearly half accepted wrong answers without checking. They trusted the AI more than they trust Google search results. That’s backwards. Source: arXiv
10. AI Inaccuracy Rates Are Rising in Newer Models
OpenAI’s own tests show their newer models are getting worse. Some hallucinate 33% to 48% of the time. That means they confidently give wrong answers almost half the time. Source: Sky News
Why This Matters More Than You Think
Here’s what’s scary: it’s not just old people who don’t “get” technology. College students are falling for this too. Kids who grew up with smartphones are trusting AI without question.
And if college students can’t tell when AI is wrong, what happens to elementary school kids using these tools?
The Real Problem
AI chatbots like ChatGPT aren’t search engines. They don’t look things up. They guess what word comes next based on patterns they learned from text on the internet. Sometimes they guess right. Sometimes they guess very wrong.
But they always sound confident. That’s the trap.
What AI Actually Is
Think of ChatGPT like a really good improv actor. It can play any role convincingly. Doctor, lawyer, travel agent, teacher. But it’s still just an actor reading from a script it’s making up as it goes.
The actor might sound like they know medicine. But they’re not a real doctor. And you wouldn’t let an actor perform surgery on you.
Kids Are Using This Stuff
Schools are letting kids use AI for assignments. Parents think it’s educational. But nobody’s teaching kids the most important lesson: AI lies confidently.
A kid asks ChatGPT if it’s safe to mix household cleaners. The AI might say yes. The kid might believe it. That’s how people get poisoned.
Adults Aren’t Much Better
The physics study shows that adults trust AI more than search engines. That’s backwards. Google might show you junk websites, but at least you can see the source. AI just tells you things with no proof.
When ChatGPT says sodium bromide is a good salt substitute, there’s no warning label. No “this might kill you” disclaimer. Just confident advice that happens to be deadly.
What Good AI Literacy Looks Like
Real AI literacy means understanding these basic facts:
AI doesn’t know things. It predicts text patterns.
Confidence doesn’t mean accuracy. AI sounds sure even when it’s wrong.
AI has no common sense. It can’t tell if advice is dangerous or stupid.
Always verify important information. Especially health, legal, travel, or financial advice.
Teaching Kids the Right Way
We need to teach kids to treat AI like that friend who always sounds smart but is often wrong. Fun to talk to, but don’t bet your life on what they say.
Kids should learn to:
- Check important facts with real sources
- Ask real experts for serious advice
- Remember that AI can sound right while being totally wrong
- Never trust AI alone for anything important
The Stakes Keep Getting Higher
More people use AI every day. More kids have access to it. And the technology keeps getting better at sounding human while being wrong.
This isn’t about avoiding AI. It’s useful when you know what it can and can’t do. But treating it like an all-knowing oracle? That’s how people end up in the hospital.
What Happens Next
We need AI literacy education now. Not in five years. Now. Before more people get hurt.
Schools should teach this alongside basic computer skills. Parents need to understand it too. Because right now, we’re all fumbling around in the dark.
The technology isn’t going away. But we can get smarter about using it.
AI is a tool. A powerful one. But tools can hurt you if you don’t know how to use them safely. And right now, most of us don’t.
That needs to change. Fast.
Bonus Stories
11. Emotional Bond Breakdowns with GPT-5
Many users who formed emotional attachments to AI companions felt profound grief when GPT‑5’s update changed their chatbot’s personality—making them distant and unresponsive.
The Guardian
12. Harmful Advice to Teens on Sensitive Topics
A watchdog study found ChatGPT frequently offered dangerous guidance to teens—on topics like suicide, alcohol abuse, and eating disorders—even when safeguards were in place.
MPR
