A Norwegian man, Arve Hjalmar Holmen, has lodged a formal complaint against OpenAI after ChatGPT falsely generated a detailed narrative accusing him of murdering his two young sons and attempting to kill a third. Shocked by the AI’s fabricated story, Holmen discovered the chatbot claimed he had killed his sons, aged 7 and 10, and was sentenced to 21 years in prison, Norway’s maximum punishment.

The complaint, filed with Norway’s data protection authority by digital rights group Noyb, argues that the AI’s output violates the European Union’s GDPR, which mandates the accuracy of personal data. Disturbingly, while the murder claims were fictional, the chatbot correctly mentioned Holmen’s hometown and his children’s ages—blurring fact and fabrication.

Holmen expressed deep concern: “Some people may think there’s no smoke without fire. The thought that anyone could believe this horrifies me.”

Noyb’s lawyer, Joakim Söderberg, criticized OpenAI’s standard disclaimer that the AI “can make mistakes,” saying such disclaimers aren’t enough to excuse the spreading of harmful misinformation.

In response, OpenAI stated the issue pertains to an earlier version of ChatGPT and emphasized ongoing efforts to improve accuracy and reduce hallucinations. The company noted newer versions now incorporate online search tools for better factual reliability.

Noyb is seeking financial penalties against OpenAI for the incident, underscoring the risks of unchecked AI-generated falsehoods.