These days, everybody’s talking about artificial intelligence or A.I. It can scour the internet and provide valuable and concise results in micro-seconds, giving the people exactly what they were looking for. Or… it can completely make things up, apparently. Here are some recent mishaps involving A.I. not having any idea what it was talking about.

• In 2026, major professional services company KPMG was tasked with researching just how deeply AI errors, falsehoods, and made-up data or hallucinations are infiltrating government documents and academic studies. When researchers published their data, it became evident that they may have also used less than reliable publicly available AI channels. In reporting on AI hallucinations, the paper cited 45 problematic incidents. Of those, five were real and 40 had been completely invented or fabricated by AI software. 

• Air Canada utilizes chat-bots on its website to help with most of its regular, straightforward customer service tasks regarding ticket purchases and changes. In 2022, Jake Moffatt used Air Canada’s chat system to find out what documents he’d need to produce to be able to qualify for a bereavement fare, or discount on a last minute flight to the site of a funeral. The chatbot told Moffatt that he had 90 days to fill out a certain form and he’d get his discount in the form of a rebate. After the travel occurred, and Moffatt applied for the discount he was rejected — Air Canada said rebates weren’t offered on travel after the fact. That’s not the policy of the airline at all, but the chatbot said it was, leading Moffatt to sue for the money owed. In 2024, he won the case.

• Chong Ke, an attorney working out of Vancouver, British Columbia, found herself under investigation by a regional oversight board in 2024 for the use of A.I. sources like chatbots in preparing legal documents. In a child custody case, Ke cited previous, precedent cases, obtained by asking OpenAI service Chat GPT. The one problem: Those cases didn’t exist in the slightest. Similarly, lawyers representing MyPillow CEO Mike Lindell in a Colorado lawsuit were each fined $3,000 when they were caught citing cases in an official filing that cited non-existent cases. 

• Cursor is a computer coding platform that utilizes AI to help its paid members write and develop software quicker. In 2025, the company tried to crack down on people in different locations sharing accounts, and announced that simultaneous logins on multiple devices would no longer be permitted. Thousands of users reportedly canceled their Cursor accounts. An internal investigation revealed that an AI-powered Cursor chatbot named Sam had told multiple users about the new login policy — which Cursor hadn’t thought about, proposed, or implemented. “Sam” wasn’t even on the pre-approved list of names used by the AI customer service program.