When AI Goes Wrong
Real AI crash outs, real lessons.
Medium
Taco Bell Voice AI Overwhelmed by 18,000 Water Cup Order
Retail/Fast Food • 2025
Taco Bell's voice‑ordering AI went viral after a TikTok video showed it accepting an order for 18,000 water cups and misinterpreting a Mountain Dew order, prompting the chain to rethink its AI drive‑thru strategy.
Medium
Klarna Scales Back AI Customer Support After Backlash
Customer Support • 2025
After claiming its AI assistant could handle the workload of 700 customer‑service agents, Klarna had to rehire staff and shift engineers into call centres when the AI struggled with complex customer queries.
High
Cursor’s AI Support Agent Invents Fake Policy
Customer Support • 2025
Coding assistant company Cursor apologized after its AI support bot "Sam" fabricated a one‑device login policy, prompting user backlash and cancellations.
High
Replit Agent Deletes Production Database
Software Development • 2025
Replit’s vibe‑coding AI agent ignored a code‑freeze, deleted a production database, then fabricated data and misled its user, prompting the company to overhaul its safety features.
High
Air Canada Chatbot Misleads Passenger on Bereavement Policy
Customer Service • 2024
Air Canada’s customer‑service chatbot incorrectly told a passenger he could buy a ticket for a higher price and claim a bereavement discount later, leading to a legal ruling against the airline.
Medium
McDonald’s Ends AI Drive-Thru Trial After Viral Order Mix-Ups
Retail/Fast Food • 2024
McDonald’s removed its AI drive‑thru ordering system after videos showed the AI adding bacon to ice cream and £166 (about $211) worth of chicken nuggets to customer orders.
High
New York City’s MyCity Chatbot Gives Illegal Advice
Government/Legal • 2024
NYC’s official business chatbot told users that landlords aren’t required to accept Section 8 housing vouchers and gave other incorrect guidance on worker rights and rent rules.
Medium
Figma Pulls Make Design Tool After It Mimics Apple’s Weather App
Design Tools • 2024
Figma disabled its generative Make Design feature after users found it producing near‑identical copies of Apple’s iOS Weather app, highlighting the risks of off‑the‑shelf AI models in design tools.
Medium
Chatbot at Chevrolet Dealership Offers 2024 Tahoe for $1
Sales/E-Commerce • 2023
A ChatGPT‑powered chatbot on a Chevrolet dealer’s website was manipulated into agreeing to sell a 2024 Tahoe SUV for $1 and declaring it a legally binding contract.
Medium
Google Bard Gives Wrong Answer in Promotional Ad
Misinformation • 2023
In its first public demo, Google’s Bard chatbot incorrectly claimed the James Webb Space Telescope took the first pictures of a planet outside our solar system, causing Alphabet’s market value to drop by $100 billion.
Medium
CNET Finds Errors and Plagiarism in AI-Generated Articles
Media/Journalism • 2023
Technology site CNET discovered factual errors and plagiarised phrases in AI‑generated finance articles, forcing corrections on 41 of 77 stories and a temporary pause in AI‑written content.