What to do when an AI chatbot gives false information about you

You ask ChatGPT what it knows about you and read that you were once convicted of fraud. Or a client sends you a Google AI Overview that mentions your company in the same breath as a scam. False information from an AI chatbot feels hard to pin down, because nobody consciously wrote that sentence. You can still act on it, with the provider of the chatbot and with anyone who passes the statement on.

Liaise acts for individuals and businesses that a chatbot says something untrue about. We draft the request to the provider, pursue whoever repeated the statement and go to court where needed.

Is a chatbot saying something untrue about you right now? Call 020 675 88 21.

Record what the chatbot says about you

A chatbot does not give the same answer every time. Tomorrow it may be worded differently or be gone altogether. So save the question you asked, the answer, the date and the name of the chatbot straight away. In ChatGPT you can also share a conversation as a link. Take screenshots of the sources the chatbot cites as well.

Then ask the question a few more times, ideally from another account too. If the false statement keeps coming back, you are in a stronger position than with one stray answer, because it shows that others who ask about you get to read the same thing. That counts in court too.

Having false information removed by the chatbot provider

OpenAI, the company behind ChatGPT, has a privacy portal. There you can ask it to stop information about you from appearing in answers, even if you don’t have an account. You prove your identity and explain why the information is wrong. OpenAI says it does not investigate disputed facts itself and weighs your request against freedom of expression and the public interest. A judgment, an entry in an official register or a newspaper’s correction is then the most convincing evidence.

OpenAI does not promise a correction. It offers to keep the information from coming back in ChatGPT, but it does not remove it from other websites or search engines. For a Google AI Overview or for Gemini you submit the request to Google.

The provider has to answer a request under the GDPR within one month. For a complex request it may take two months longer (article 12(3) GDPR).

Your GDPR rights when a chatbot gives false information about you

When a chatbot says something about you as an identifiable person, the provider is processing your personal data. That data must be accurate (article 5(1)(d) GDPR). You have the right to rectification of inaccurate data (article 16 GDPR) and in many cases a right to have it erased (article 17 GDPR). If you suffer harm, you can claim compensation (article 82 GDPR).

With a language model it is not yet clear how those rights play out. In 2024 OpenAI told the privacy organisation noyb that it could not correct inaccurate data in ChatGPT, only block it for certain prompts. In March 2025 noyb filed a complaint on behalf of a Norwegian man about whom ChatGPT had invented that he had murdered two of his children. That complaint now sits with the Irish data protection authority, because OpenAI Ireland Limited is responsible for European users. There is no decision yet.

If the provider rejects your request or does not respond, you can complain to the Dutch Data Protection Authority. For providers whose European establishment is in Ireland, such as OpenAI and Google, the Irish authority has the final say. That takes time. You can also ask the court to order the provider to grant your request after all, within six weeks of its answer (article 35 of the Dutch GDPR Implementation Act). You bring those proceedings in the Netherlands, even against a provider based in Ireland (article 79(2) GDPR).

What German courts decided about Grok and Google’s AI Overview

There is no Dutch ruling yet on false information from a chatbot. Germany has two. On 23 September 2025 the Hamburg court prohibited xAI from spreading, through the Grok account on X, the claim that the campaign group Campact lived largely on government money. For the court it made no difference that Grok had generated the text itself. Users read such an answer as fact, all the more so because Grok itself says its answers are based on facts. xAI had to answer for the statement because it appeared on its own account (LG Hamburg, 324 O 461/25).

In May 2026 the Munich court went a step further, in a case about a Google AI Overview that linked two publishers to fraud. Google argued that as a search engine it only shows information from others. According to the court, the AI Overview summarises the search results in its own words, so Google makes statements of its own and is responsible for them (LG München I, 26 O 869/26). Both decisions were given in urgent proceedings. The Munich decision is not yet final.

Dutch law works with the same idea. Here too you can be liable in tort for someone else’s statement if you adopt it and spread it (article 6:162 of the Dutch Civil Code). That applies all the more to a provider that generates the text itself. Such a provider also cannot rely on the protection the Digital Services Act gives platforms for what their users post. That protection only covers information provided by a user (article 6 DSA).

When a chatbot misrepresents your business

The GDPR only protects people. If it concerns your company, your foundation or your brand, the route is tort law, because a business also has a right to protect its reputation. The Munich case was about two publishers.

Alongside the legal steps, look at where the chatbot gets its information. A chatbot that searches the web often names its sources. The error is frequently already there, in an old news item, a namesake or a wrong listing in a business directory. Have that source corrected or removed, so that the chatbot does not land on the same error the next time someone asks. See also what you can do about reputational harm to a business.

When someone spreads a chatbot’s false statement

It often doesn’t stop at the answer itself. Someone posts a screenshot on LinkedIn, a journalist copies a generated summary or a competitor forwards the answer to clients. Each of them turns the statement into their own publication. Even when the text comes from an AI, whoever spreads it has to check first whether it is true.

You can demand that such a person takes the publication down and publishes a rectification. For a rectification, the person spreading it need not even have known the information was wrong (article 6:167(2) of the Dutch Civil Code). The court may then make you share the costs. If the other side doesn’t cooperate, interim injunction proceedings follow. Defamation comes into play as well if someone spreads the accusation knowingly.

How to demand a rectification →

Lawyer for false information from an AI chatbot

We work for people a chatbot misrepresents and for businesses that an AI Overview confuses with fraud. We draft the request to the provider with the evidence it needs, send a cease-and-desist letter to whoever spread the statement and ask the court for an order or an injunction where needed. Because no Dutch court has yet ruled on such a case, we tell you upfront what is achievable and where it remains uncertain. If you use AI to publish yourself, read what AI law means for you.

Is a chatbot saying something untrue about you or your business? Get in contact or call 020 675 88 21. You will have an answer within one working day.