A man from Norway has lodged a formal complaint against the entity responsible for ChatGPT, following an incident where the chatbot inaccurately accused him of murdering his own children.
Arve Hjalmar Holmen, describing himself as an ordinary individual without any public presence in Norway, sought information about himself from ChatGPT and was met with a distressing response. The chatbot alleged that Holmen was a father who had brutally killed his two sons, aged seven and ten, in an incident that occurred near his home in Trondheim back in December 2020, adding that the man was sentenced to 21 years in prison for the heinous crime.
In his complaint, filed with the Norwegian Data Protection Authority in collaboration with Noyb, a group advocating for digital rights, Holmen pointed out that though the fabricated account had Bhar NSLog elements that mirrored his real life—such as his hometown, the number and ages of his children—the story was entirely false and could have dire consequences on his personal life if leaked within his community.
The complaint asserts that Holmen has never faced accusations or convictions of any criminal activity and is regarded as a law-abiding citizen. The complaint also contends that ChatGPT’s error breached the GDPR’s requirements for data accuracy and has requested OpenAI to modify its model to prevent similar inaccuracies and impose a sanctions on the company.
Following Holmen’s interaction with ChatGPT, OpenAI released an updated model with internet search capabilities, which has reportedly made it less likely to reproduce the error in the future. AI chatbots, including ChatGPT, operate by predicting likely word sequences, making them susceptible to providing false and nonsensical responses that may appear accurate to users.
A spokesperson from OpenAI stated that the company is actively exploring ways to enhance the accuracy of its models and diminish “hallucinations,” or false outputs. While reviewing the complaint, the spokesperson acknowledged that the incident involved an older version of ChatGPT and pointed to the model’s recent upgrades that improve its accuracy.
Source: https://www.theguardian.com/technology/2025/mar/21/norwegian-files-complaint-after-chatgpt-falsely-said-he-had-murdered-his-children
