A German court just did what years of AI ethics panels, congressional hearings, and think-piece essays couldn’t: it held Google directly liable for false statements its AI generated.

Not indirectly. Not “should have caught it.” Directly responsible — as the author of the content.

The Munich Regional Court ruled on May 28 that Google’s AI Overviews aren’t just search results with a fancy wrapper. They’re Google’s own words. And when those words are wrong — defamatory, fabricated, confidently false — Google owns the consequences.

If this holds up, it doesn’t just change the game for Google. It rewrites the legal playbook for every AI company shipping products right now.

The Case: AI Hallucinations Meet Real Victims

Two Munich publishers discovered that when users Googled their companies, AI Overviews cheerfully informed them these businesses were “known for dubious business practices” and “often perceived as a scam.”

None of it was true.

Google’s AI had done what LLMs do best and worst: it confused these companies with other businesses flagged for shady behavior, then served up the mashup with total confidence. The AI even opened with authoritative declarations that no human source had ever made.

Google’s defense? The AI warns users information might contain errors. People should verify things themselves.

The court’s response? Essentially: that’s not how any of this works.

The Court Dismantled Every Big Tech Defense

The ruling (case no. 26 O 869/26) systematically torched the arguments Google — and by extension, every AI company — has been leaning on.

AI Overviews aren’t search results. Traditional search displays links to third-party content. AI Overviews synthesize, rephrase, and generate new statements from multiple sources. That’s not indexing. That’s authorship. The court drew this line with surgical precision.

The AI made things up. The defamatory claims didn’t appear in any of the underlying search results. Google’s AI fabricated connections that no human source had ever made. This wasn’t amplification — it was creation.

Only Google can fix it. Since Google built, trained, and deployed the system, it bears sole responsibility. The original websites never made these claims, so victims can’t sue them. Without holding Google liable, victims have “virtually no legal recourse.”

That last point is the killer. The court refused to accept a world where AI confidently destroys reputations and nobody’s accountable.

The “Don’t Trust It” Defense Is Dead on Arrival

Google’s go-to argument — users know AI can make mistakes, they can click through to verify — ran into a wall of data.

The court cited Pew research showing that when AI Overviews appear, clicks to traditional results drop from ~15% to ~8%. Only about 1% of users click a source link from an AI Overview. People read the summary and move on. They’re not verifying anything.

The judges also spotted the obvious contradiction: if Google itself says AI Overviews “should not be blindly trusted,” then the feature’s utility “would be significantly diminished.” Why does it exist if you’re telling people not to believe it?

This creates what the court called a “protection gap.” The AI makes confident statements. Users treat them as gospel. Nobody checks. And if Google isn’t liable, nobody is.

Algorithms Don’t Have Opinions

In a move with far-reaching implications, the court also addressed whether AI output deserves free speech protection.

The answer: no.

An AI’s “opinion” is “not the expression of an acquired conviction” but “the result of an algorithm.” Deploying AI-powered search is “above all an expression of Google’s business activities,” not free expression.

This directly counters arguments in US courts where chatbot makers have claimed AI output constitutes protected speech under the First Amendment. The Munich court’s position: algorithms don’t have opinions. Companies have products. Products have liability.

Why Every AI Company Should Care

This is a German court ruling — preliminary, and Google will appeal. But the legal logic is platform-agnostic.

The reasoning transfers everywhere. If Google’s AI Overviews are Google’s “own statements,” then ChatGPT’s responses are OpenAI’s own statements. Perplexity’s answers are Perplexity’s own statements. Microsoft Copilot, same deal. Any system that synthesizes information into new content falls under this framework.

The scale of exposure is terrifying. Google processes over five trillion searches annually. A New York Times analysis found AI Overviews are accurate roughly 90% of the time — which means hundreds of millions of potentially wrong answers per year. Over half of “accurate” responses were “ungrounded,” meaning cited sources didn’t actually support the claims.

Disclaimers won’t save you. Building a system designed to be trusted, then disclaiming responsibility when it’s wrong? The court said that’s not a legal shield. It’s a contradiction.

The Liability Reckoning Was Always Coming

This ruling lands at a moment when the AI industry is sprinting in the opposite direction from caution. OpenAI just partnered with Visa for AI-powered purchases. Agents are being deployed across enterprise workflows. The entire industry is racing to make AI more authoritative, more trusted, more autonomous.

But authority comes with accountability.

If an AI agent books a flight based on wrong information, who pays? If an AI health chatbot gives dangerous advice, does a disclaimer protect the company? Munich says no.

We’re watching two simultaneous reckonings. In the US, the government is forcing Anthropic to take Fable 5 offline over capabilities concerns. In Germany, courts are tackling the other end: AI that’s not capable enough but deployed as if it were.

Both cases point to the same truth: the “move fast and break things” era of AI is colliding with “you break it, you buy it.”

What Happens Next

Google will appeal. Legal experts expect this could reach Germany’s Federal Court of Justice or influence EU-level policy under the AI Act.

But the signal has been sent. The “it’s just AI, don’t blame us” era is ending. Companies deploying AI at scale need to think harder about accuracy, liability insurance, and what happens when their systems confidently state things that aren’t true.

For the rest of us, the takeaway is simpler: that confident AI summary at the top of your search results might be making things up. And finally, in at least one courtroom, someone’s being held accountable for it.