FOLLOW US

AI hallucinations are a well-documented phenomenon in artificial intelligence (AI). Large Language Models (LLMs) like ChatGPT generate text by predicting the most probable next word in a sequence, but they do not inherently understand the context. 

When they fail to find an appropriate response, they sometimes generate misinformation—what is commonly referred to as an AI hallucination.

An AI model hallucinates when it provides false or misleading information. While this might seem harmless or sometimes even amusing, the consequences can be severe, particularly when AI-generated misinformation influences healthcare, legal matters, or personal reputations.

Confused and Falsely Accused

One alarming instance of AI hallucination involved Norwegian citizen Arve Hjalmar Holmen. He searched his name in ChatGPT to see what information OpenAI’s chatbot would provide.

To his horror, ChatGPT falsely claimed that he had murdered two of his children, attempted to kill a third, and had been sentenced to 21 years in prison.

The Austrian-based European privacy advocacy group, None of Your Business (NOYB), took up Holmen’s case, filing a 2025 complaint with Datatilsynet, Norway’s Data Protection Authority. 

NOYB investigated the origins of the fabricated claims, checking if anyone with a similar name had committed serious crimes. Their search yielded nothing, reinforcing that the chatbot's AI model had fabricated the allegations.

What made the situation even more alarming was that ChatGPT mixed these false claims with accurate personal details about Holmen, including the names and genders of his children and his hometown.

 While OpenAI has since updated the AI model, removing the false information, NOYB raised concerns about whether the erroneous data remains embedded within the LLM’s dataset.

How AI Hallucinations Occur

The concept of hallucinations takes on a different meaning in AI.

 According to Geeks for Geeks, “AI hallucination refers to AI systems generating imaginative novel, or unexpected outputs. These outputs frequently exceed the scope of training data.”

AI hallucinations occur when an AI model generates inaccurate answers that either do not belong to the training data or are entirely fabricated. 

This often happens when the AI perceives patterns or objects that do not exist. In some cases, AI tools produce outputs incorrectly decoded by transformers, leading to incorrect information being presented as fact.

The Risk of Persistent Misinformation

NOYB warns that AI hallucinations may not be fully correctable. “The incorrect data may still remain part of the LLM’s dataset. By default, ChatGPT feeds user data back into the system for training purposes. This means there is no way for the individual to be absolutely sure that this output can be completely erased [...] unless the entire AI model is retrained.”