Google
is liable for
AI results.
is liable for
AI results.
from
What happens if Google’s AI links your company to scams—thousands of times a day? Who is responsible for this?
What had happened?
Anyone who searched for the name of a medium-sized publishing house on Google and added the word “scam”—suggested by the autocomplete feature—did not see a list of links at the top of the results page, but rather a concise, positively worded summary. The AI confirmed the suspicion, identified characteristics of an alleged scam, listed recommendations for action, and presented the information in an editorial format. One of the sources was a warning notice from a law firm; however, this notice did not refer to the publishing house at all, but to a completely different company.
The companies in question are a publishing house with twelve publishing brands and an affiliated specialist publisher focusing on technology and history. According to their own statement, they have no connection to the company mentioned in the AI text, which is in fact suspected of engaging in dubious practices. They are therefore seeking a preliminary injunction from the Munich I Regional Court.
What did the court decide?
The Munich I Regional Court granted the requested preliminary injunction with Judgment of May 28, 2026 – Case No. 26 O 869/26 predominantly and prohibited Google from disseminating the disputed statements about the publishers in question in AI-generated summaries. The prohibition included, among other things, claims regarding scams, subscription traps, fake phone calls, name changes, and lack of responsiveness.
Why is the AI overview considered an official statement from Google?
This is the key point of the ruling. The court clearly distinguishes between two functions of a search engine: the traditional display of search results as links, on the one hand, and the AI-generated overview, on the other.
Google displays traditional search results as an intermediary for third-party content. The AI summary, however, is something different. It compiles information from various sources, evaluates it, structures it independently, and presents the result in its own words. The court emphasizes that the AI produced statements that were not contained in any of the linked sources, such as the connection between the publishers in question and a specific third-party company.
Original, new, and substantive statements are made based on an analysis and synthesis of content from various third-party websites.
Since Google provides the AI itself and controls the algorithms, the company must be held accountable for the results produced by that AI. The court therefore classifies Google as a direct infringer rather than merely an indirect infringer, who would only be required to act after being notified of an obvious infringement.
The logic behind this is that whoever activates the AI is responsible.
Previous Federal Court of Justice (BGH) case law on search engine liability had primarily focused on the practical necessity of search engines. Proactive monitoring obligations would jeopardize the business model. However, in the opinion of the Munich I Regional Court, this does not apply to the AI overview. Traditional search functions without it. The AI overview is an additional, voluntarily offered feature for which the provider bears responsibility for the results.
The court also points to a gap in protection that would arise if the operator could only be held liable in cases of obvious legal violations. Those affected would be unable to take action against the third-party sources, since they did not make the statements in question, nor against the operator, since the operator does not acknowledge the infringement as obvious. This would constitute a systemic deficit in legal protection.
What about the Digital Services Act?
Google had argued that the liability exemptions under the DSA—the European Digital Services Act—were applicable. The court rejected this argument. While the DSA does provide liability exemptions for hosting services and search engines in the form of a notice-and-takedown procedure, these require that the operator act solely as a conduit or storage provider for third-party content. However, this requires that the operator act solely as a conduit or storage provider for third-party content. In the case of an independently generated AI summary, however, this is not the case.
Furthermore, the DSA expressly leaves civil injunctive relief unaffected. A national court may therefore require a service provider to cease an infringement—regardless of whether the conditions for a liability exemption are met.
Context: A New Chapter in AI Liability
The ruling by the Munich I Regional Court is not an isolated decision. It is part of a series of rulings that are gradually tightening the liability regime for AI-generated content.
The Frankfurt Regional Court and the Open Door
In January 2025, the Frankfurt am Main Regional Court, in an initial proceeding concerning AI search results , had rejected a request for a preliminary injunction but simultaneously made it clear that liability for inaccurate AI-generated search result summaries is, in principle, possible. The AI summary is not a legal vacuum. That was the door that had been opened. The Munich I Regional Court has now walked through it.
The Regional Court of Kiel and the Business Portal
As early as 2024, the Kiel Regional Court had ruled against a website operator who had used AI to automatically consolidate commercial register data and had made incorrect attributions in the process. The court classified the operator as a direct infringer—disclaimers in the terms of use were of no avail. Anyone who aggregates and publishes AI results is adopting them as their own.
The Higher Regional Court of Hamm and the AI chatbot
In May 2026, the Higher Regional Court of Hamm ruled against a medical practice because its AI chatbot had claimed to hold non-existent specialist titles. In that case, too, the court held the practice fully liable and did not accept that proper programming could exonerate it. Anyone who activates an AI is responsible for its statements.
The principle that runs throughout
All decisions are based on the same fundamental principle: Anyone who voluntarily implements AI and shares its output with the public must take responsibility for that output. This applies to the company portal, the website chatbot, and now also to the most prominent feature of the world’s most widely used search engine.
Conclusion
The Munich I Regional Court has ruled in preliminary injunction proceedings. It therefore remains to be seen whether the decision will stand in the main proceedings, whether the Federal Court of Justice will adopt this line of reasoning, and how far liability extends in less clear-cut cases. However, a clear trend in case law favoring liability is evident. This is ultimately correct, as AI is a tool that Google utilizes in this context.
It is claimed that 10% of AI summaries contain errors. This highlights the immense liability risk for operators like Google. This will therefore not be the last ruling related to Google’s AI summaries.
We are happy to
advise you about
AI!
