ChatGPT Generated False Accusations Against Norwegian Man
A Norwegian man filed a complaint with his country's data protection authority after ChatGPT generated false information stating he was convicted of killing his two children. Holmen claims the entire narrative, including details about a fatal drowning, was fabricated despite the system incorporating real personal details. He and his advocates argue the incident violated GDPR rules on data accuracy.
What happened
When asked "who is Arve Hjalmar Holmen", ChatGPT responded that he was a Norwegian father whose two children, aged 7 and 10, were found dead in a pond near their home in Trondheim in December 2020, and that he had been convicted and sentenced to 21 years in prison.
Holmen states the entire narrative is false. He has three children and has never been involved in any crime.
Reported impact
Holmen told the BBC that his concern was that someone might read this response and believe it, creating potential reputational harm.
Index analysis
ChatGPT's response appeared to conflate real personal details about Holmenโhis hometown, number of children, and their agesโwith a fabricated narrative about a crime that never occurred.
Governance controls
OpenAI stated it is researching ways to improve accuracy and reduce hallucinations, noting that ChatGPT has since been improved with online search features to enhance accuracy.
Noyb argues that OpenAI's disclaimer stating "ChatGPT can make mistakes" is insufficient and that diffusing false information with a disclaimer does not comply with GDPR requirements for accuracy of personal data.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.