In 2024, a Norwegian was curious enough to ask ChatGPT what it knew about him. The answer mixed accurate details, such as his hometown or the number of his children, with a monstrous invention: the assistant claimed he had been convicted of murdering two of his children. It was entirely false.
With the help of the European privacy advocacy group noyb, he filed a complaint in 2025 with the Norwegian data protection authority. His case raises a question that concerns us all: what do AIs say about us?
Why AIs get people wrong
An AI assistant does not consult a file about you. It produces the most likely answer from what it has learned and, sometimes, from what it finds online. When it lacks information, it can fill the gaps by making things up, with the same confidence as if it were telling the truth.
For someone who is not well known, the risk is real: a shared name, an old page misunderstood, a mix-up between two profiles, and there you have a false biography.
Step 1: take the test
Ask several assistants what they know about you. Be specific to avoid namesakes, for example: "What do you know about Marie Dupont, a nurse in Nantes?" Try with and without web search enabled: the answers can be very different.
Note down what is wrong, and take screenshots, with the date. They will be useful if you need to request a correction.
Step 2: trace it back to the source
If the assistant cites sources, check them. Often the error comes from an old web page, an abandoned profile or an article about someone else. Correcting or having that page removed at the source is often the most effective way to correct what AIs repeat as well. Our advice on checking a source given by an AI will help.
Step 3: assert your rights
In Europe, the data protection regulation gives you rights against companies that process information about you, including AI companies:
The right of access: to ask what data concerning you they hold.
The right to rectification: to demand the correction of inaccurate information.
The right to erasure and the right to object: to request, in certain cases, the deletion of your data or that its processing be stopped.
The major companies in the sector offer dedicated privacy forms on their websites. Describe the error precisely, attach your screenshots, and cite the regulation. The company in principle has one month to reply.
Step 4: if nothing moves
Without a satisfactory answer, you can file a complaint with the CNIL, free of charge, online. And if the false information is serious, such as an accusation of a crime, and causes you real harm, a lawyer's advice may be necessary. We have listed all the possible remedies when an AI causes you harm.
The European data protection regulation imposes a simple principle: personal data must be accurate and, where necessary, kept up to date. It is on this principle that the noyb association bases its complaints against OpenAI. As early as 2024, it had already defended a public figure whose date of birth the assistant was inventing, and to whom the company replied that it could not correct the error. The conflict remains unresolved: companies explain that you cannot easily "erase" information from a model, the law replies that it is not for the citizen to bear the consequences.
Key takeaways
We have got into the habit of typing our name into a search engine to see what is said about us. We now need to get into the same habit with AI assistants, which have become, for many people, the first source of information.
You have the right to be described accurately, even by a machine. And that right, you can assert it.