The Munich Regional Court I issued a preliminary injunction on May 28 that AI companies have spent months dreading: Google must answer for statements its Gemini chatbot produces, even when they are wrong. The court found Google directly liable for hallucinated answers delivered through its AI-powered search features, a first in Europe, according to lawyers involved in the case.
What makes the ruling unusual is that Google had done the things platforms usually do to protect themselves. The AI answers carried source links. The interface carried a disclaimer that responses “may be wrong.” The court said none of it matters, and that is the part of the decision the rest of the industry is now reading closely.
The court’s reasoning is direct. Google markets Gemini as a fast way to get reliable information, so it cannot turn around and tell users not to trust the results it presents. The company can’t have it both ways: either the answers are reliable enough to stand behind, or the feature shouldn’t be presented as a source of information at all.
The underlying dispute involved a user who said Gemini’s answer was factually wrong and caused him harm. The court did not need to resolve the factual dispute to issue the injunction; under German procedure, a preliminary order requires a plausible claim and urgency, and the court found both.
The holding reflects a structural problem with AI search. When a chatbot fabricates a citation, the “third-party source” it names often doesn’t exist or doesn’t say what the model claims. The victim has no one else to sue, so the responsibility comes back to the platform that produced the answer. The Munich court is the first in Europe to say so explicitly.
German law makes the ruling harder to escape than an ordinary product-liability decision. Courts here have long held platforms responsible for content they present as their own, and the preliminary injunction is enforceable while Google appeals, meaning the company must change behavior now and sort out the merits later.
The decision lands as regulators across Europe are still building the enforcement machinery for the AI Act, the bloc’s sweeping rules for artificial intelligence. The Act imposes transparency obligations on general-purpose models, but it is a framework, and the Munich ruling shows national courts moving faster than the regulators who are supposed to implement it.
Google said it is reviewing the ruling and stands by its safeguards, which include source links and disclaimers, the very features the court rejected. The company’s argument, that it merely assembles information from across the web, collides with a court that sees the assembled product as Google’s own statement.
The ruling is not the first time a court has wrestled with chatbot speech. Courts in Australia and Canada have allowed defamation and libel claims against chatbot makers to proceed, and the German decision extends that line into the core of search, where AI answers now appear above the links that used to be the entire product.
Lawyers said the Munich ruling could become a template. The court’s logic, that advertising reliable information while disclaiming it is contradictory, applies to any chatbot with a search interface, and German courts routinely influence litigation in other European jurisdictions. A company that loses one case in Munich faces dozens of copycat claims.
For Google, the practical response is limited. It can appeal, it can tighten the disclaimers, or it can change what Gemini answers. None of those options restores the old position, where a search engine pointed at sources and let users judge. The court has effectively said that once Google starts answering, Google owns the answer.
The wider implication is for every company shipping consumer AI in Europe. The disclaimers that labs bolt onto their products, the “this is AI-generated” labels, the “may be wrong” footnotes, all of it is now weaker than the marketing that surrounds it. If a company advertises its assistant as a way to get information, a court may hold it to that promise.
The Munich order is preliminary, and the full case will take years.
The ruling also lands amid a lobbying fight. Tech companies have argued in Brussels that AI providers shouldn’t be treated like publishers of every output, and the Munich decision cuts against that argument. Industry lawyers said the case will be cited in every regulatory hearing about AI liability for the next year, and that European courts have a habit of borrowing each other’s reasoning.
The injunction, unless overturned on appeal, requires Google to change how Gemini presents answers in Germany. The company has options: add a verification step, flag answers as unverified, or remove the AI answer box entirely for queries where it can’t guarantee accuracy. Each option carries a cost, and the court’s framing leaves little room for a middle position. But the direction is set: for two years, AI companies have answered hallucination complaints with disclaimers, and a Munich court just told the industry that the disclaimer is the problem, not the solution. The appeal will be watched from Palo Alto to Brussels.
