Munich Court Rules Google Directly Liable for False Claims Made by Its AI Overviews

Munich Court Rules Google Directly Liable for False Claims Made by Its AI Overviews

Category: Policy | The Vault — AI Edition

A German court has done what no court before it appears to have done: hold a company legally responsible for what its AI says. In a ruling that is already reverberating through Europe's tech sector, the Regional Court of Munich (Landgericht München) found Google directly liable for false statements generated by its AI Overviews search feature — and pointedly rejected the company's core defense that AI summaries are merely algorithmic output users should know to double-check.

The decision, issued as a temporary injunction in case no. 26 O 869/26 and dated May 28, 2026, bars Google from repeating false claims about two Munich-based publishing companies through its AI-generated search summaries. According to reporting by German outlet The Decoder, which obtained the ruling, the court classified Google as a "direct infringer" because, in its view, the AI Overview is Google's own content — not a neutral list of links pointing elsewhere.

What the court found

The dispute began when Google's AI Overviews allegedly tied two publishers to scams, subscription traps, and "dubious business practices" in response to certain search queries. The catch, according to the court: the accusations appeared in none of the sources the AI cited. The system had, in effect, invented the connections, blending information about genuinely sketchy companies with the plaintiffs who had no link to them. The publishers, supported by sworn affidavits, sent a cease-and-desist letter; the court found Google did not respond appropriately.

The court's reasoning turned on a distinction that traditional search law has never had to confront. A conventional search engine, German courts have long held, merely makes third-party content findable and therefore enjoys limited, "indirect" liability. AI Overviews, the Munich judges concluded, are different in kind. The AI rewrites and judges results "in its own words and according to its own structure," the ruling states, producing "independent, new, and substantive statements." Because those statements are Google's own, the court reasoned, "it alone has influence over the AI's offering and the algorithms with which the AI operates."

Crucially, the court rejected the safe-harbor logic that has shielded search engines under rulings from Germany's Federal Court of Justice. "Only Google can check those statements," the court noted, "at least by comparing the underlying third-party websites with its own statements based on them." Host-provider protections under the EU's Digital Services Act, the judges added, did not apply.

"Nobody needs AI to search the internet"

The most quoted line from the ruling captures its skepticism toward Google's framing of the feature as essential. The court characterized AI Overviews as "an additional function — one without which the use of the search engine would still be (and is) possible, and without which users are perfectly capable of finding results amidst the 'flood of data.'" The widely circulated paraphrase: nobody needs AI to search the internet.

The implication is sharp. Because the AI Overview is presented as an authoritative, self-contained answer rather than a list of sources, the court held it to a higher duty of accuracy — not a lower one. Google's defense, that users "can dig deeper and verify" the linked sources, fell flat. The possibility of disproving a statement through further research does not "regularly exempt from liability for this statement," the court ruled, drawing a parallel to press law, where publishers are liable for self-contained teasers even if readers never click through. The point lands harder given research from Pew showing roughly 1 percent of users ever click a source link inside an AI Overview.

Google's response

Google has signaled it will appeal and disputes the characterization of its product. "We invest deeply in the quality of AI Overviews to ensure that the overwhelming majority of responses provide accurate information, and they are designed to reflect the information that exists on the web," a Google spokesperson said in a statement provided to The Decoder. "We're carefully reviewing this decision, which is not yet final." The company confirmed in mid-June that it intends to challenge the ruling.

That the decision is a temporary injunction, not a final judgment, matters. But the court ordered Google to cover 80 percent of the legal costs and found a continuing risk of repeat violations, since nothing technically stops the algorithms from generating the same claims again.

The transatlantic divergence

The ruling's significance extends well beyond two German publishers. It crystallizes a widening gap between how Europe and the United States treat AI-generated speech. In the US, Section 230 of the Communications Decency Act has historically immunized platforms from liability for content — though whether that shield covers a company's own generative AI output remains an open, hotly contested question, with several US defamation suits against AI firms still unresolved.

The Munich court charted the opposite course. By declaring that an AI's output is the operator's own statement rather than passively surfaced third-party content, it imports publisher-level liability standards that simply do not currently apply to AI search in the US. For European operations of Google, OpenAI, Anthropic, and Perplexity, the message is that AI-generated content presented as fact may be judged against defamation law — a standard the American market has so far largely avoided.

The stakes are not hypothetical. An analysis by AI startup Oumi, reported by The New York Times, found Google's AI Overviews answered correctly about 91 percent of the time using its current Gemini 3 model — impressive, but at Google's scale still millions of wrong answers an hour. The same analysis found that 56 percent of even the correct answers could not be traced to the linked sources. That untraceability is precisely what the Munich court seized on.

What to watch

Three threads will determine how far this ruling travels. First, the appeal: a higher German court could narrow or overturn the reasoning, and Google has every incentive to fight a precedent that, if upheld, could force stronger guardrails and explicit legal responsibility onto AI answer engines across the EU. Second, the interaction with the EU AI Act, now in phased implementation through 2026 and 2027. The Munich court's instinct — to assign responsibility to the deployer of an AI system rather than treat its outputs as neutral content — aligns closely with the Act's emphasis on accountability, suggesting the two regimes may reinforce one another.

Third, contagion. As Reporters Without Borders put it, the principle that generative AI cannot be held to account "has begun to crack." If other European courts adopt the Munich logic, the liability exposure will not stop at Google. Every provider whose system paraphrases the web as authoritative fact — ChatGPT, Claude, Perplexity included — could find that, in Europe at least, the AI's words are now the company's words.

Reporting in this article draws on the published ruling and coverage by The Decoder, Malwarebytes Labs, The Business Standard, and Reporters Without Borders. Specific case details are as reported; the injunction is not yet final and is under appeal.