Skip to content
News Regulation

German court rules Google is liable for false claims in AI Overviews, stripping search-engine safe harbor

· by Pondero Newsdesk

The short version

The Regional Court of Munich issued a temporary injunction against Google on May 28, 2026, ruling that AI Overviews constitute Google's own editorial speech and that standard search-engine liability protections do not apply.

German court rules Google is liable for false claims in AI Overviews, stripping search-engine safe harbor

The Regional Court of Munich issued a temporary injunction against Google on May 28, 2026, ruling that the company is directly liable for false statements generated by its AI Overviews feature and that existing search-engine safe-harbor protections do not cover AI-generated content.

What happened

The court's ruling (case no. 26 O 869/26) followed a complaint by two Munich-based publishers whose reputations were damaged by AI Overviews that wrongly associated them with scams, subscription traps, and shady business practices, per The Decoder's reporting on the ruling. According to the court, the AI mixed up information about genuinely problematic companies with the plaintiffs and drew connections that did not appear in any of the linked source pages. The publishers issued a cease-and-desist letter; Google did not respond adequately.

The court classified AI Overviews as Google's own editorial content rather than an index of third-party material. It found that the AI had generated "independent, new, and substantive statements" that evaluated and combined content from multiple sources, producing claims "that are not even made in the search results." The ruling document, published by The Decoder, quotes the court directly: because Google alone controls the AI and its algorithms, Google alone bears responsibility for what the AI produces.

Google was ordered to stop displaying the specific false claims and to cover 80 percent of legal costs. The plaintiffs each pay 10 percent. The court noted that the risk of repeated violations remained because Google had not issued a cease-and-desist declaration with a penalty clause, leaving the algorithms free to regenerate the same statements.

On June 11, after the original ruling became public, Google issued a statement saying its AI Overviews "are designed to reflect the information that exists on the web" and that the company is "carefully reviewing this decision, which is not yet final," per The Decoder. The statement also repeated the argument the court had already rejected: that users can verify claims by consulting the linked sources.

Why the court rejected Google's defenses

Three of Google's arguments were dismissed in the ruling, and each refusal has implications for how courts may treat AI-generated content going forward.

First, Google argued that existing German Federal Court of Justice (BGH) precedent shielding search engines from liability should apply. The Munich court disagreed. BGH precedent rested on the premise that search engines merely make third-party content findable and that a proactive duty to check results would undermine how search works. AI Overviews do not merely link. They generate new text, structure it under headings, and state conclusions. That changes the legal category, the court found.

Second, Google argued that users know AI-generated content can be wrong and can check sources themselves. The court called that insufficient. The AI Overview was "understandable on its own" and contained "a self-contained statement with independently understandable content." The possibility of disproving a false statement through further research does not exempt the author from liability for that statement. The court cited a Pew study finding that only 1 percent of users click a source link directly from Google's AI Overviews as evidence that the practical user experience is to accept the overview at face value.

Third, Google raised Digital Services Act host-provider protections. The court declined to apply them, finding that Google does not passively host third-party content in AI Overviews. It originates the content.

The ruling also addressed free speech. AI-generated text does not represent an acquired conviction of any person, the court wrote. It is algorithmic output in service of a commercial product. That framing gave the plaintiffs' privacy rights priority over any speech interest Google could claim.

Why it matters for AI answer engines

The ruling's scope extends past this single dispute. An analysis by AI startup Oumi, cited by the New York Times and referenced in The Decoder's report, found that Google's AI Overviews with the Gemini 3 model answered correctly 91 percent of the time. At Google's scale, the remaining 9 percent still amounts to millions of wrong answers delivered every hour. If enough of those errors defame individuals or organizations, the Munich ruling provides a legal template for holding Google accountable for each one.

The court also noted the ruling may have international reach. That is not a binding declaration, but it signals that the Munich court considered the precedent implications. Other EU member-state courts could cite this ruling when evaluating similar complaints under their own defamation and press-law frameworks.

For publishers and businesses concerned about how AI answer engines describe them, this ruling matters in a specific practical way. It establishes that the operator of an AI overview cannot point to source pages and disclaim responsibility. The AI's summary is the operator's statement. If that statement is false and damaging, the operator faces direct liability in at least one German court.

The pattern also applies beyond Google. Services that generate synthesized answers from web sources, including ChatGPT search, Perplexity, and Claude's web-browsing responses, face the same structural question: when the AI says something the sources do not say, who is responsible? The Munich court's answer is the operator. Whether courts in other jurisdictions reach the same conclusion is open, but the German ruling is now a point of reference for plaintiffs and regulators across the EU.

What to watch next

Google described the ruling as "not yet final," signaling a likely appeal. The first thing to track is whether a German appeals court (Oberlandesgericht) upholds, narrows, or overturns the liability classification. A sustained ruling from a higher court would significantly strengthen the precedent.

The second thing to watch is regulatory response. The EU's AI Act includes provisions on high-risk AI and transparency obligations, and regulators in Brussels and Berlin have already signaled interest in how AI Overviews handle accuracy. If the Munich ruling stands, it gives regulators a domestic-court basis for pushing broader compliance requirements onto AI answer engines operating in EU markets.

Third, watch for copycat claims. The Munich publishers' case succeeded partly because the AI's error was traceable and the harm was specific. Any organization that has been mischaracterized by an AI Overview now has a German ruling to cite as a jurisdictional model.

Sources