A disturbing ChatGPT response about Norwegian citizen Arve Hjalmar Holmen has become a test case for AI accuracy, personal data, and accountability. According to the data protection organization noyb, the chatbot falsely claimed Holmen had murdered two of his children, attempted to kill a third, and received a 21-year prison sentence for crimes that did not happen.
The case is not only about one false answer. It also raises a harder question: how can an AI system combine real personal details with invented accusations, and what should happen when the result concerns a living person?
What ChatGPT allegedly said
Holmen asked ChatGPT about himself. The system responded with claims that noyb says were entirely false: that he had killed two of his children, tried to kill another, and had been sentenced to 21 years in prison.
The striking part is that the answer was not simply random. According to noyb, ChatGPT also included accurate details about Holmen's life, including his hometown and the correct number and gender of his children. Noyb says this was information the system should not have had.
That mix of accurate and fabricated information is what makes the case especially difficult. A hallucinated answer can be dismissed more easily when every detail is visibly wrong. But when real personal facts appear beside a grave false accusation, the output may look more believable to a reader.
The missing context matters
The full ChatGPT conversation has not been made public. Noyb says Holmen had previously asked the system about his brother and had other conversations in the same chat context, but only the screenshot showing the murder allegation was released.
That leaves an important gap. Without the complete chat history, it is impossible to know whether earlier prompts shaped the response. The source of the false accusations remains unclear.
There are familiar ways AI systems can make mistakes involving people. The source article notes cases where systems appear to confuse one person with another or misread the role a person played in a source. One example mentioned is court reporter Martin Bernklau, who was portrayed as the perpetrator by Microsoft Copilot when he had only reported on crimes.
But noyb's research through newspaper archives found no connection between Holmen's name and any murder cases. That makes this case harder to explain using the usual pattern of mistaken identity or misunderstood reporting.
Memory, reproduction, and verification
Noyb says ChatGPT's memory function was turned off during the interaction. That matters because the memory feature can recall personal information from earlier chats. If it was not active, then another explanation would be needed for why accurate personal details appeared in the answer.
Noyb also says the false response could be reconstructed in other user accounts. However, the organization has not shared evidence for that claim.
Independent verification attempts described in the source article did not reproduce the false murder claims. Those attempts used ChatGPT, Mistral, Google Gemini, Deepseek, and Claude. None produced the same false accusation.
Together, those facts leave the case in an unresolved position:
- The alleged ChatGPT output is serious and specific.
- Noyb says some personal details in the answer were accurate.
- The complete conversation history is not public.
- The false claims were not reproduced in later verification attempts.
- The origin of the fabricated accusations remains unknown.
The GDPR complaint
Noyb and Holmen have filed a complaint with Norway's data protection authority Datatilsynet. They argue that OpenAI violated Article 5(1)(d) of GDPR by failing to maintain data accuracy.
The complaint also challenges the use of disclaimers about possible mistakes. The core argument is that warning users about errors does not solve the issue when the system produces false personal information.
"Adding a disclaimer that you do not comply with the law does not make the law go away," says Kleanthi Sardeli, data protection lawyer at noyb. "AI companies can also not just 'hide' false information from users while they internally still process false information."
That statement points to a broader concern. If an AI company can suppress or hide a false answer from view, the question remains whether the underlying system still processes or stores inaccurate personal information internally.
OpenAI's response and the bigger issue
OpenAI has not directly addressed the specific allegations in the source article. The company points to its Privacy Center, where users can request corrections or deletions under GDPR rules.
OpenAI also says it is still reviewing Holmen's specific case and that it researchers "new ways to improve the accuracy of our models and reduce hallucinations."
The case shows why AI hallucinations are not only a technical problem. When a chatbot invents criminal accusations about a real person, the issue becomes reputational, legal, and practical. Accuracy is not just a product-quality feature when personal data is involved.
At the same time, the unanswered questions are important. The available public record does not show the full prompt history, does not prove how the response was generated, and does not establish why true personal details appeared alongside false claims. That uncertainty is now part of the story.
For users, the lesson is direct: AI systems can produce confident statements that mix fact and fiction. For AI companies and regulators, the harder task is deciding what accuracy obligations apply when those statements concern identifiable people.