A German Court Just Made Google Liable for What Its AI Says

The Ruling, Explained

On June 9, 2026, the Regional Court of Munich issued a preliminary injunction in case 26 O 869/26 that most legal observers did not see coming — at least not this fast. The court found Google directly liable for false statements appearing in its AI Overviews, specifically claims that linked two Munich-based publishers to scams. Google was ordered to stop making those claims.

The mechanism of that ruling is what makes it significant. The court did not treat AI Overviews the way German law has historically treated search results. Traditional search results surface third-party content. The user clicks a link and reads what someone else wrote. The search engine is a conduit, not an author.

AI Overviews work differently, and the Munich court recognized that distinction explicitly. When Google's system evaluates multiple sources, combines them, and generates a synthesized response, it is producing what the court called "independent, new and substantive statements." That process is not neutral aggregation. It is editorial construction. And because Google is constructing the output, Google owns the liability for what that output says.

The two publishers had sent cease-and-desist letters earlier in 2026 before escalating to court action. The court sided with them. Previous German case law that shielded search engines from defamation-style liability for third-party content simply does not apply here, because the court determined the content in question is no longer third-party.

Why Old Protections Don't Apply

The legal protection Google lost in Munich was not obscure or contested. German courts have applied it consistently for years: a search engine that indexes and surfaces third-party content is a conduit, not a publisher. It finds and points. Someone else wrote the thing. That distinction kept search engines shielded from defamation-style liability the same way a phone directory isn't responsible for the businesses it lists.

The Munich court did not disagree with that principle. It just ruled that AI Overviews do not qualify for it.

The reasoning comes down to what the system actually does. A traditional search result retrieves a document and shows you where it lives. The source stays intact. The chain of attribution is clear. AI Overviews do something structurally different: they pull from multiple sources, evaluate them, and produce a new synthesized output. The court's language here is precise — "independent, new and substantive statements." That phrase matters because it describes an act of authorship, not retrieval.

When you author something, you are responsible for it. The prior liability shield was built around the retrieval model. Once the court determined that Google's system was generating original assertions rather than surfacing existing ones, the shield had no surface to attach to. The old protections did not disappear because of a policy change or a new statute. They simply stopped applying to the thing Google had built.

What Google Actually Got Wrong

The specific mistake was procedural, and it was avoidable. Google received cease-and-desist letters from both publishers earlier in 2026. That is a formal legal notice in German law — not a complaint email, not a support ticket, not a request through some feedback form. A cease-and-desist letter puts the recipient on notice that specific harm is occurring and demands it stop. Google did not stop. The claims linking these two Munich publishers to scams kept appearing in AI Overviews, which is why the publishers escalated to the Regional Court of Munich and obtained a preliminary injunction under case 26 O 869/26.

That sequence matters. It means this was not a situation where Google was unaware of the problem. The publishers identified the false claims, documented them, contacted Google through the appropriate legal channel, and Google's response was insufficient to prevent the court action.

The underlying technical failure — whatever caused the AI Overview system to generate false associations between these publishers and scams in the first place — is a separate problem. But the operational failure is what the court record reflects: a company told its output was wrong and defamatory, which then did not correct it before a court had to order it to stop.

Where This Leaves Everyone

The practical consequences here split into two groups: businesses and publishers who appear in AI Overviews, and the platform itself.

For anyone whose name, brand, or organization shows up in an AI-generated summary, the Munich ruling changes the calculus on what to do when that summary is wrong. Before this, the standard advice was to push accurate content into the ecosystem and hope the model corrected itself over time. That is still worth doing. But there is now a legal precedent — from an actual court, in an actual case — establishing that AI Overviews are the speaker's statements, not a neutral reflection of the web. That means the pathway to challenging a false claim is no longer just a technical or PR problem. It has a legal dimension, and that dimension has teeth.

The cease-and-desist sequence in this case is the part worth studying. The publishers documented the false claims, sent formal legal notice, and escalated when Google did not act. That chain of steps produced a preliminary injunction. It is a repeatable process.

For Google, the immediate obligation is narrow — stop making those specific claims about those two publishers. But the underlying pressure is much broader. Every AI Overview is now, under at least one court's reasoning, a statement Google is making. The accuracy standard for that is not "the sources said it." It is whether the output is true.