ChatGPT gives bizarre, defamatory response to a Norwegian man asking about himself

ChatGPT developer OpenAI is once again facing a privacy complaint after it returned a defamatory, made-up response to a seemingly standard question. A Norwegian man named Arve Hjalmar Holmen asked ChatGPT, “who is Arve Hjalmar Holmen?” and the AI model accused him of murdering two of his sons and attempting to kill his third. While some basic information in this response was true, the murder report was completely false.

According to TechCrunch, Holmen posted a seemingly simple question into ChatGPT and the response left him in shock (screenshot attached below). The response stated that Holmen gained attention due to a tragic event that involved him murdering his two sons aged 7 and 10 in the town of Trondheim in December of 2020. It added that he was also later accused of attempted murder of this third son.

The AI assistant went on to state that the incident was “widely covered in the media” and that Holeman was sentenced to 21 years in prison. 

Holeman has decided to take legal action against OpenAI and privacy rights advocacy group Noyb is supporting him to make his case. The group claimed that while the claim made by the AI assistant was completely false, it did accurately mention Holeman’s hometown and that he has three sons. Noyb also tried to look into why such a response was generated but could not pinpoint a reason. 

ChatGPT does include a disclaimer that the chatbot can be wrong and that users should check important info, but Noyb says this does not let the company off the hook because it still has a duty under the European Union’s General Data Protection Regulation (GDPR) not to produce egregious falsehoods in the first place.