- Lead. Munich’s Regional Court has ruled that Google is directly liable for false statements its AI Overviews made about two local publishers, stripping the company of the platform-immunity protections that have shielded search engines for decades.
- Fact. The court, in case 26 O 869/26 decided on May 28, found that AI Overviews generate “independent, new, and substantive statements” rather than listing third-party content, making Google a primary author rather than a conduit.
- Stake. The ruling’s logic applies to every AI answer engine — including ChatGPT, Claude, and Perplexity — and could expose them all to defamation claims across EU jurisdictions that follow the Munich court’s reasoning.
The case arose after Google’s AI Overviews wrongly linked two Munich-based publishing companies to scams, subscription traps, and dubious business practices, apparently by mixing up their information with that of genuinely problematic firms and drawing inferences that appeared in none of the cited source pages. The publishers sought a temporary injunction, which the Regional Court of Munich granted on May 28, barring Google from continuing to circulate the false claims and ordering it to cover 80% of legal costs.
Why traditional immunity did not apply
Google argued that its AI Overviews should be treated as a listing of third-party content — the framework under which search engines have been insulated from liability under both US Section 230 and EU host-provider rules since the 1990s. The court rejected that argument on three grounds.
First, AI Overviews do not merely reproduce what source pages say; they synthesise, paraphrase, and draw inferences, producing output that is formally new. Second, because “Google alone has influence over the AI’s offering and the algorithms,” the court classified the company as a direct infringer rather than a passive intermediary. Third, the court dismissed the argument that users could fact-check claims themselves, noting research showing that users rarely follow the source links that AI Overviews display.
The accuracy paradox at scale
One of the ruling’s most pointed observations involves scale. Even a hypothetical 91% accuracy rate translates into millions of erroneous statements at Google’s query volume. The court also found that 56% of the system’s correct answers could not be traced to the sources cited, meaning the underlying retrieval mechanism itself generates hallucinated attribution. The Digital Services Act’s host-provider protections were found inapplicable to AI-generated summaries, according to the ruling, because AI Overviews represent an additional function layered on top of search rather than search itself.
That distinction echoes the logic European regulators have applied in related areas — the same separability principle has underpinned Digital Markets Act enforcement requiring Meta to open WhatsApp’s AI features to rivals, where bundled AI functionality is treated as separable from the core platform.
Implications beyond Google
The ruling is a first-instance decision, subject to appeal, and Google has not stated publicly whether it will contest it. But the precedent it establishes — that AI answer layers constitute their author’s speech rather than a reflection of underlying search results — is likely to travel quickly across German courts and to influence how other European jurisdictions frame the liability question.
For the industry, the decision arrives as AI Overviews, AI Mode in search, and AI-generated summaries across platforms are becoming the primary interface through which hundreds of millions of users encounter information. If the Munich court’s reasoning holds, any AI system that synthesises sources and presents conclusions as informational statements faces a fundamentally different liability exposure than the search engines they have been built on top of.