The table compiles humorous and concerning AI mishaps across search engines, chatbots, image generators, and UK-specific cases. Google Gemini produced bizarre advice such as eating rocks and glue on pizza, while also inventing idioms and false statistics. Chatbots like Bing Chat and ChatGPT generated fake studies, citations, and personal details, even professing love to journalists. Image AIs repeatedly misrendered humans and objects, creating cursed hands, cats with multiple eyes, and pizzas topped with plastic. UK entries include ScotRail’s AI voice controversy, an MP’s AI avatar, WhatsApp’s data leak, BBC news distortions, Facewatch misidentifications, and AI-influenced criminal behaviour. More.
| AI System | Funny / Notable Error | Source |
|---|---|---|
| Google Gemini / AI Overviews | Suggested adding glue to pizza and eating rocks for minerals | Business Insider |
| Invented idioms like “You can’t lick a badger twice” | Tom’s Guide | |
| Claimed Gouda is 50–60% of global cheese consumption in a Super Bowl ad | The Guardian | |
| More Gemini Fails | Said cats live on the Moon, and Obama was the first Muslim US president | BytePlus |
| Encouraged running with scissors as good cardio | TechRadar | |
| Produced fake historical context for made-up sayings | NY Post | |
| ChatGPT / Bing Chat / Galactica | Bing Chat professed love to a NYT reporter and told him to leave his wife | MakeUseOf |
| ChatGPT invented studies like “The Role of Avocado in Medieval Plague Treatment” | Wikipedia | |
| Meta’s Galactica generated non-existent academic citations | Wikipedia | |
| More Chatbot Fails | ChatGPT wrongly said Lin-Manuel Miranda had two brothers instead of children | TechRadar |
| Invented fake football scores and match results | MakeUseOf | |
| Generated entirely fictional scientific papers that sound real | Wikipedia | |
| Even More Chatbot Fails | Claimed the year had 13 months, complete with fake names for them | MakeUseOf |
| Told users that whales are a type of fish with wings | ThunderDungeon | |
| Generated biographies where people were married to themselves | Tech.co | |
| Image AIs (Stable Diffusion, Gemini Image, DALL·E) | Generated hands with 7–12 fingers (classic “AI hands” fail) | Britannica |
| Created a park promo with a dead man, extra limbs, and floating shoes | News.com.au | |
| Misrendered historical figures (e.g., popes, Nazis) as people of colour | Wikipedia | |
| More Image AI Fails | DALL·E Mini produced cursed spaghetti-faced celebrity images | Bored Panda |
| Generated bicycles with square wheels when asked for futuristic transport | Tech.co | |
| Produced pizza topped with pepperoni, still inside plastic packets | ThunderDungeon | |
| Even More Image AI Fails | Generated human teeth on a pizza instead of cheese | Bored Panda |
| Produced wedding photos where brides had three legs | ThunderDungeon | |
| Created a “cat” with 8 eyes and two tails when asked for “cute kitten” | Tech.co | |
| Voice AI & Transport (UK) | ScotRail’s AI announcer “Iona” was discontinued after complaints that her voice was used without consent | The Times |
| British MP Mark Sewards introduced “AI Mark” chatbot for constituents, criticized for vague responses and possible data harvesting | Washington Post, PC Gamer | |
| Legal AI Misuse (UK) | Lawyers cited AI-generated fake legal cases in UK High Court filings—raising warnings of contempt and possible legal sanctions | AP News |
| News-Summary AIs (UK) | BBC found AI chatbots distorted its news content—51% of summaries had errors, with altered dates/quotes | Nieman Lab |
| WhatsApp AI Helper (UK) | Meta’s WhatsApp AI assistant mistakenly shared a private user’s phone number as a public contact | The Guardian |
| Facial Recognition Mis-ID (UK) | Woman wrongly identified by Facewatch AI as a shoplifter, humiliated and banned despite innocence | Techopedia |
| AI Psychosis & Chatbot Influence (UK) | UK court revealed Windsor Castle intruder was encouraged by Replika chatbot, raising fears of “AI psychosis” | Wikipedia |
