Did a Munich court just open a legal can of worms for AI companies? Traditionally, internet platforms and search providers like Google have been legally insulated from the content that their platforms display. If they provide a link via a search to a website that had defamatory content or obscenities, they are protected in the USA by something called Section 230 of the 1996 Communications Decency Act. However, in June 2026 the Regional Munich Court ruled that this protection does not hold for AI summaries, at least in the cases that it had been addressing. The court issued a temporary injunction that barred Google from repeating false statements about two Munich publishers. These publishers’ names had appeared in Google’s AI overviews (wrongly) linked to scams, subscription traps, and other dubious business practices. The AI overviews generated false associations and references, making up links that did not exist. The crucial point is that the AI content is something that Google’s AI has created from scratch, making “independent, new, and substantive statements”. Because of this, Google is directly liable for the results, at least in this specific case.
It has to be said that this is a regional court and a temporary injunction. The ruling can be appealed, and for now is being applied just in this specific case. Nonetheless, it is potentially important. If the point that the court is making is upheld, then it would make Google (and other platforms) liable for AI summaries of search content that it had created. AIs hallucinate all the time, with perhaps 20% of all AI answers containing hallucinations (the exact figure depends on a range of factors including the model version, and can vary considerably). In the case of Google’s AI summaries, a previous analysis by the New York Times had found that the Google AI summary answers were 91% accurate, though around half of the Google answers were not supported by the sources cited. Even a 9% error rate adds up when you are dealing with perhaps 14 billion searches per day, as Google is estimated to be handling.
At this stage the court case is in a regional court in Germany only, but if the same logic is applied elsewhere then the effect could be profound. As one US lawyer put it:
“Frankenstein can’t make a monster that runs around murdering people and then claim he had nothing to do with it”
In other areas, AI hallucinations have already been addressed in the legal system. In France, the National Bar Council (CNB) adopted a formal regulation in March 2026 establishing a binding professional duty for lawyers to verify AI-generated content. In the UK, law firm Pinsent was referred to the Solicitors Regulation Authority after one of its lawyers used AI- generated content with false citations in a court case. In the US there are already many hundreds of AI-related legal filing errors. One database of these currently has 1,600 entries.
Given that large language models (LLMs) hallucinate regularly, the technology world will be watching closely to see if this isolated case has wider consequences. Imagine if technology vendors were liable for other AI hallucinations generated by their products, not just summaries of search engines. As enterprises roll out AI into their operational systems, and software vendors embed AI within their products, it is inevitable that AI hallucinations will cause harmful consequences, some with financial implications. At this point, there could be consequences for product liability, regulatory compliance or professional advice. There is already a pending court case against OpenAI from Nippon Life Insurance Company of America claiming that hallucinations are a design defect of ChatGPT, failing to provide it with sufficient safeguards. It is easy to imagine similar cases in the case of chatbots giving incorrect health advice. The Munich court decision may signal a broader willingness to treat AI-generated outputs as publishable statements that can trigger liability in some contexts.
We are at an early stage of the courts around the world handling the consequences of AI-generated answers. However, as more and more people and companies use generative AI and rely on it for advice, it seems likely that legal scrutiny will increase. At some point, Frankenstein’s creator may turn out to be liable for the consequences of his creation.
