2026-06-10
Landmark German ruling declares Google’s AI Overviews are Google’s own words and makes it liable for false answers by Matthias Bastian.
A German court has ruled that Google is directly liable for what its AI search overviews say. Previous case law shielding search engine operators from liability doesn’t apply to AI overviews.
I love to see it. Of course it’s difficult to imagine this type of ruling happening in the United States.
Google’s AI overviews had falsely tied two publishing companies to scams, subscription traps, and shady business practices for certain search queries. According to the court, the AI mixed up information about other, genuinely sketchy companies with the plaintiffs and drew connections that didn’t appear in any of the linked sources.
There’s existing case law that provides limited exceptions (“host provider protections under the Digital Services Act” ?) for search engines and autocomplete, since by their nature a search engine is surfacing statements made and published by others. The court’s reasoning in this case focuses on the fact that AI summaries are, basically, original content created by the technological system itself. In this case specifically, the AI summaries included claims that were not supported by any of the related search results, but I think the court’s argument applies regardless of the accuracy of the summaries.
David Chisnall makes the point that Google directly argued in the case that it’s common knowledge AI summaries are not reliable, and that by entering this into evidence, it allows others to rely on this statement from Google in future lawsuits.
The interesting thing about the German court ruling against Google is not the verdict. The fact that, if you put libel on your web site, you are liable for it even if you used a machine to automatically generate libel, should not surprise anyone who has paid attention to the law at any point in the last century or so: humans have agency, the tools that they use do not shield them from liability, no matter how obfuscating they are.
The bit I suspect will have much more impact longer term is one of the defences entered by Google’s lawyers. Somewhat more verbose in the original German, but it boiled down to: Everyone knows LLMs produce nonsense, no one should ever trust the output of an LLM in any situation that matters, it’s not Google’s fault if people read the output of an LLM and believed it might have some connection to reality.
It’s debatable whether everyone knows that, but this is now an official statement entered into the court record that at least one of the major LLM vendors knows this. And that’s now an on-the-record statement made under penalty of perjury that can be entered as evidence in any court case against companies selling LLM-integrated tooling.
I suspect that this will show up in a lot of court cases over the next few years and probably have a much bigger long-term impact than the ruling. Any claim about utility made by vendors of ‘AI’ tools is now open to lawsuits ranging from misleading advertising to outright fraud as a result of this.
Google would probably have been much better advised to settle the case rather than enter that claim as evidence. Imagine if a car manufacturer had entered a defence against liability in case of a collision by saying ‘everyone knows automobiles are impossible to operate safely on the roads and anyone who buys one should know better than to take it on the public highway’. Google’s lawyers have just done the equivalent for the ‘AI’ industry.
Addendum 2026-06-25. AI and Liability - Schneier on Security by Bruce Schneier summarizes this better than I did, and perfectly matches my point of view on these systems.
Google’s AI overviews are far less nuanced. They work differently from traditional search, which courts have held involves archiving and facilitating access to the editorial content of third parties. AI overviews don’t just quote and republish words from different websites. With overviews, the AI rewrites other people’s words, exercising editorial discretion like a newspaper article or an original essay on a topic.
Section 230 has never made sense to me. One hundred authors submit essays to the New Yorker, it (the corporation) exercises its editorial discretion over which to publish, and it must stand liable for those decisions. One million users on Facebook write comments, Facebook exercises (algorithmically mediated) editorial discretion over which of those comments to publish to me, and may disclaim any liability. Insanity.