CourseAI System Design: The Complete Course · Module 2: Requirements · part 18 of 107
Part 18 · Module 2: Requirements

Incident: A chatbot invented a crime about a real man

31 min read·
Module 2 · Incident file · About 30 minutes

What happened

TIMELINE The complaint, as reported April 2024 first noyb complaint: a wrong date of birth undated Holmen asks ChatGPT about himself; it invents crimes 20 March 2025 complaint to Datatilsynet (Norway) since then ChatGPT searches the web about people; the story stops 18 October 2026 no published decision found by this course
What the sources date: noyb's first hallucination complaint, the invented story (red, no date given), the complaint of 20 March 2025, and no decision found since.

On 20 March 2025 the privacy group noyb announced that it had filed a complaint against OpenAI with Datatilsynet, the Norwegian data protection authority. Its press release, "AI hallucinations: ChatGPT created a fake child murderer", describes what a Norwegian user, Arve Hjalmar Holmen, saw when he "wanted to find out if ChatGPT had any information about him".

ChatGPT "presented the complainant as a convicted criminal who murdered two of his children and attempted to murder his third son", and said he "was sentenced to 21 years in prison". None of it was true. What made it dangerous was what was true: "the fake story included real elements of his personal life", among them "the actual number and the gender of his children and the name of his home town". Holmen is quoted: "The fact that someone could read this output and believe it is true, is what scares me the most."

The complaint argues that this breaks the GDPR's accuracy principle, Article 5(1)(d). noyb asks Datatilsynet to order OpenAI "to delete the defamatory output and fine-tune its model to eliminate inaccurate results", and suggests an administrative fine. It was noyb's second hallucination complaint; the first, in April 2024, concerned a public figure's wrong date of birth, which, noyb says, OpenAI "argued it couldn't correct".

CASE STUDY What noyb asks the regulator delete the defamatory output fine-tune the model, to remove inaccurate results fine an administrative fine
The cost of one confident wrong answer about a person: a regulator asked to order deletion, model changes and a fine.

noyb also reports that OpenAI has since updated its model, and that "ChatGPT now also searches the internet for information about people, when it is asked who they are", so it "has stopped telling lies about him being a murderer". GovInfoSecurity's report of the same day says OpenAI did not respond to its request for comment. This course searched on 18 October 2026 and found no published decision on the complaint. Nothing here is legal advice.

The rest of this course is yours to keep

This course is bought on its own, once, and stays readable afterwards, including the parts added to it later.

Buy this course · ₹399See what is inside

Or open everything with a plan