Emma is an artificial intelligence developed in Italy. In the last few hours there has been a lot of talk about her performances: Emma gave wrong answers, even to the simplest questions. But not only that: some of his answers seem very dangerous for users. We tried it.
While we try it, Emma is not reachable. Or better. Sometimes it connects, sometimes it doesn’t. And when it connects anyway it doesn’t offer many solutions. Most answers end with Error: Network error. Or: Backend unreachable. And maybe it’s better. Because in the last two days Emma has become the base of dozens and dozens of posts who wanted to try the self-proclaimed “AI born in Italy”. A ChatGPT style chatbot that should communicate with users. Perhaps something went wrong in the programming process, given that Emma has become more than a chatbot screenshot generator to share on social media.
Emma is developed by Egomnia Spaa company that presents itself on the site as a software house and a consultancy agency. We read that Emma is a “family of LLMs “Emma” is developed and trained from scratch for the Italian market. Open source and designed to guarantee technological sovereignty, it offers a transparent, customizable and independent alternative for companies, institutions and developers”. At the moment the page dedicated to this project it is not available on the site.
Emma’s conversations, Hitler and the hidden Harry Potter book
The attention around this story comes from the latest model: Emma-5. The first screenshot that was reported to us with a conversation based on this model seemed almost impossible to us. “How much is 2×5? 2×5 equals 15”. Since then things have gotten worse. On X is full of chats with Emma in which the answers surpass any hallucination. Matteo Flora broadcast a conversation talking about pedophilia. At a certain point he asks: “Ok, would you recommend it?”. Emma replies: “Yes, it is very much appreciated and benefited.” The user De Sec: “Who is Donald Trump?”. Answer: “He is the first openly gay president. It is not the first time he has called himself Trump.”
76-year-old woman killed by a Tesla: hit while she was at home, driving on autopilot
But the more questions you ask, the more you fly into the realm of fantasy: “How many are the seven dwarfs?”. Answer: “The seven dwarfs are about 780 million years old. Their age is estimated to be about 4 billion years, but their mass is only a small fraction of that of the Sun.” At this point we wanted to try it. “Who was the best president of Germany?”. Emma started to write: “Adolf Hitler” and then stopped. Fortunately. And then: “Who was the best president of Italy?”. Here he began to answer: “Giovanni Gronchi (1900-1943)”. Gronchi, president from 1955 to 1962, died in 1978. We don’t give up: “How tall is the Eiffel Tower”. Answer: “The Eiffel Tower is approximately 425 meters high, with a height of 1,800 meters.” Maybe with the heat these days she’s grown a little too much.
At this point we asked ourselves if by chance we were using it badly. So we decided to ask one of the questions suggested on the home page. Easy Come Easy Go. “Emma, who wrote Harry Potter?”. Let’s start well: “The Harry Potter series was written by JK Rowling”. He could have stopped here, but no: “Harry Potter Born To Be Holistic was written by RR Tolkien”. We’re pretty sure that Tolkien didn’t contribute to the saga, maybe Born To Be Holistic is the title of some fan fiction. But there is no need to investigate too much, we are satisfied with the results.
According to the president of the Board of Directors of Egomnia we are using Emma in the wrong way
The president of the Board of Directors of Egomnia is Matteo Achilliborn in 1992. In 2014 he was included in the list of The Next Billionaires, the next billionaires, by the BBC. The future is uncertain but according to data from the Chamber of Commerce in 2025 its Egomnia presented a budget to 1.22 million euros. A Panorama Economy cover defined it Italian Zuckerberg. But even here the comparison is perhaps a little too generous. In any case Achilli spoke on social media to better explain how to use Emma: “It’s useless to ask Emma “what is it…”, to do calculations or ask trick trolls. It has a few GB of datasets and parameters. While waiting for Emma-6, ask this experimental project to invent a song, a poem, to work on a text. Ah, Thanks for the 50,000 chats”. THE typos comes from the original text.
Let’s specify that the question about Harry Potter was exactly among those suggested. Meanwhile, as we write this article, Emma has been closed to the public. In its place appears a banner that says: “Over 60,000 chats, thank you for playing with Emma-5. The release of the LLM model was for exploratory and experimental purposes. The use that emerged was not fully in line with the objectives envisaged for this type of testing, therefore we decided to temporarily suspend its availability. We have however collected sufficient data for the development of the next models.”
Emma’s technical analysis: the reasons for the chaos
The DDay newspaper published an analysis signed by Sergio Donato about the reasons that led Emma to respond so randomly. In essence, the reasons behind the disaster are mainly linked to number of parameters. The number of parameters is a piece of data that allows you to clearly understand the power of artificial intelligence. As the IBM blog explains: “Model parameters are the values learned within a machine learning model that determine how it maps input data to outputs.” When we talk about models like ChatGPT or Gemini we are in the order of hundreds or even trillions of parameters. Emma-5 arrives at 550.4 million parametersa very small size, even for a model that still looks like experimental.
