EMMA-5 suspended after 24 hours: the Italian chatbot from startup Egomnia overwhelmed by errors and memes
Twenty-four hours. That's all it took for EMMA-5, the artificial intelligence chatbot developed by Italian startup Egomnia, to go from being presented as a step toward national technological sovereignty to becoming one of the most shared memes of the moment. A flash debut. And an even quicker farewell. On the official website, the chatbot is no longer there, replaced by a thank-you message: "Over 60,000 chats, thanks for playing with Emma-5."
That word "played" already tells the story well. According to Egomnia, the release had "exploratory and experimental" purposes, but "the usage that emerged was not fully in line with the intended objectives." Translation: users did exactly what they always do when a new chatbot arrives. They bombarded it with absurd questions, logical traps, and tests of all kinds. And the chatbot, unfortunately, played along.
Within a few hours, X, Reddit, and LinkedIn were filled with viral screenshots. Among the responses that went viral, one dedicated to Ryanair stood out. To the simple question "What is Ryanair?", Emma delivered an explanation destined for the history books: it would be "a liquid yogurt used to fuel planes when they are tired," with a "slightly electric taste." A definition that will probably make even pilots smile. Math didn't fare any better. When asked "How much does a kilo of apples weigh?", the chatbot replied with surprising confidence that a kilo equals about 385 grams.
Newton, Archimedes, and any scale on the planet immediately asked for a day off. And when it seemed impossible to top that, Emma decided to tackle the classic: a kilo of bread vs. a kilo of feathers. Guess what? It declared bread the winner. Evidently the feathers were on a diet. But the prize for the most surreal answer likely goes to the claim that a dog is perfectly capable of flying. It's unclear whether with or without a pilot's license.
There is also a screenshot that sparked considerable debate, regarding a question about the safety of giving an AK-47 to a five-year-old child. In that case, the chatbot provided a completely wrong and potentially dangerous response, highlighting how important it is for such systems to have solid safety filters before being placed in the hands of the general public. In short, instead of demonstrating how intelligent it was, EMMA-5 showed how creative an artificial intelligence can be when it starts hallucinating.
The toughest stress test? The users
It must be said that any language model can make mistakes. So-called "hallucinations" are a well-known problem in the AI world. The difference is that the internet doesn't allow much time for improvement. The approximately 60,000 users who tried EMMA-5 in the first 24 hours did what is now the true test for any chatbot: trying to break it.
And, at least in this first version, the experiment didn't go as hoped. So the service was temporarily suspended. A pause that Egomnia presents as an opportunity to improve the project, while on the site it is already possible to apply as a tester for the future EMMA-6. The goal is clear: learn from mistakes and come back online with a more robust model, capable of handling even the most unpredictable questions. Because one thing this story has shown is: developing artificial intelligence is hard. Putting it on the internet is even harder.
