Liability
for AI
AI search.
Liability
for AI
AI search.
from
Google and other search engines now display AI-generated summaries above the classic search results. What happens if these summaries contain incorrect information? The Frankfurt am Main Regional Court ruled for the first time: Liability is possible in principle – but denied in the specific case because the challenged statement was not false in the overall context.
When AI hallucinates: Who is responsible?
Imagine a potential customer searching for your service offering and the AI search at the top of the results page warns of life-threatening risks that don’t even exist. A nightmare for any company. But are search engine operators actually liable for the nonsense they generate with the help of algorithms and artificial intelligence in the prominent “AI Overview” box? Or do Google & Co. enjoy the freedom of fools when the machine “hallucinates”?
The new era of search results
The days when a search engine simply provided a list of links are over. With the introduction of “AI Overviews”, search engine operators generate direct answers to user questions. This is convenient for the user, but is explosive for companies: The AI summarizes information, interprets it and – here’s the catch – sometimes gets it wrong.
As these AI boxes often take up the entire screen area “above the fold” (the visible area without scrolling), the information presented there becomes the ultimate truth for the casual reader. If this is wrong, there is a risk of massive reputational damage.
A medical procedure and a “severed” truth
A controversial case in the field of aesthetic surgery ended up before the Frankfurt am Main Regional Court. A medical service provider defended itself against a representation in the AI overview of the search engine giant.
The point of contention was information on a penis thickening operation. In its summary, the AI claimed that the penis was “severed” during the procedure. In fact, however, the procedure only involves cutting the retaining ligament in order to achieve an optical lengthening.
The medical service provider understandably saw this as a false claim that was damaging to its business and would massively deter patients, and demanded an injunction and applied for a temporary injunction.
The verdict: No preliminary injunction
The Regional Court of Frankfurt am Main dismissed the application for a temporary injunction with Judgment of 10.09.2025 – Ref. 2-06 O 271/25 but at the same time created important legal guidelines for the future.
The court ruled in favor of the search engine operator, but not because it was untouchable per se.
Liability of the search engine operator for incorrect information that was generated by an artificial intelligence (AI) system before the search result is not excluded per se.
At the same time, however, the judges stated:
However, undue obstruction can be ruled out if the challenged statement is not false according to the understanding of the addressees.
Why the AI gets away with it (this time)
The court’s reasoning is extremely informative in practice and is based on three pillars:
No free ride for search engine operators
The court clarified that AI-generated content is not a legal free zone. If obvious falsehoods are spread, there may be a claim for injunctive relief. There is no immunity for the platforms’ own content.
The “intelligent user” reads the context
In the specific case, the search engine operator was saved by the users’ reading skills. The court argued that, in the overall context of the AI search, the reader understands that the organ is not amputated in the event of a “thickening”. Although the wording was imprecise, it was not misleading enough for the average user to justify an infringement of competition law.
Antitrust law & DMA (Digital Markets Act)
The service provider had also tried to argue via the Digital Markets Act (DMA). This prohibits gatekeepers (such as Google) from giving preference to their own services. However, the court did not see the AI overview as a separate “product” that competes with the plaintiff’s website, but as an integral part of the search. The AI does not direct the user to a competing team of doctors from Google, but merely summarizes information.
Conclusion
The Regional Court of Frankfurt am Main has opened the door for the liability of search engine operators for AI search results, even if it has not yet been breached in this specific case. Search engine operators cannot claim that “the AI did it”.
Anyone who is massively harmed by AI hallucinations has – with the right arguments – a good chance of successfully defending themselves.
We are happy to
advise you about
AI!
