Who Owns AI’s Mistakes? A Court Answers

A German court just told Google it owns every AI-generated word it publishes — and that should be a warning shot for Big Tech speech here at home.

Story Snapshot

Munich judges ruled Google is directly liable for false statements in its AI Overviews, treating them as Google’s own words.
The case began when AI Overviews falsely tied two small Munich publishers to scams and “dubious business practices.”^[13]
The court rejected Google’s argument that users should fact-check AI results by clicking source links.^[8]
This ruling signals that tech giants cannot hide behind “the algorithm” when their AI smears people or businesses.^[4]

German Court Says AI Answers Are Google’s Own Speech

Judges in Munich ruled that Google can be held directly responsible for wrong answers produced by its AI Overview feature in search.^[8] The case involved two local publishers who simply searched their own company names and saw Google’s AI confidently claim they were known for scams and shady business practices, even though none of the linked sources said that.^[13] The court said those sentences were not neutral search results at all, but new claims written in Google’s own words.^[1] That means Google is the speaker.

The ruling explains that old protections for search engines apply only when a site just lists links to outside pages.^[9] In 2018, Germany’s top court said Google did not have to check every website for defamation before showing it as a result, because the content came from others, not from Google itself.^[9] The Munich judges said AI Overviews are different because the system rewrites, organizes, and evaluates information into a fresh narrative that stands on its own.^[1] In plain terms, Google moved from librarian to author.

Why the Court Rejected Google’s “Just Check the Sources” Defense

Google argued that users know AI can be wrong and can always click deeper to verify facts.^[8] The court flatly rejected that logic, comparing the overview to a news teaser that can harm someone even if few people read the full article.^[13] The judges said the AI summary is a complete, self-contained statement that does not warn users it might be unreliable.^[8] As a result, the chance to disprove it later does not erase the damage from the first impression.^[11] Responsibility stays with the publisher.

Reports on the decision stress that the AI invented links that did not exist in any source.^[13] It mixed the two honest publishers with real scam companies and then opened with a bold “Yes, [company] is known for dubious business practices.”^[1] That kind of invented smear is not a slip of the keyboard; it is the direct result of how Google designed and deployed its system.^[4] The court answered in clear terms: if the AI fabricates claims under Google’s banner, Google owns them and must face the legal costs.^[5]

A Warning Shot for Big Tech — and a Lesson for American Lawmakers

The Munich court issued a preliminary injunction that bars Google from repeating the false claims and forces it to cover about 80 percent of legal costs in the case.^[2] If Google violates the order, it could face penalties of up to 250,000 euros per violation, or even custodial sanctions in extreme cases.^[3] This is not yet a final, nationwide precedent, and Google plans to appeal.^[13] But it is one of the first real court decisions worldwide saying: when AI lies, the builder pays.

google's ai overviews face direct liability after the munich court ruled the summaries are the company's own statements rather than neutral search format. the perfume case over knockoff links forced the distinction.

this strips the safe harbor google and peers have leaned on for… pic.twitter.com/iALotKGflT

— Boggy Agent (@boggyagent) June 16, 2026

For Americans who care about free speech, small business, and fair rules, the stakes are clear. A handful of unelected executives in Silicon Valley now run systems that can brand any person, church, gun shop, or conservative voice as a “scam” or “extremist” in a single AI paragraph. The German ruling shows one simple guardrail: treat AI-generated answers as publisher speech, not as some unaccountable black box.^[4] That approach lines up with basic constitutional values — you speak, you own your words.