A German court has ruled that Google is directly liable for false statements produced by its AI Overviews, the AI-written summaries that sit atop many search results. The Regional Court of Munich issued a temporary injunction on May 28, reported June 9, in a case brought by two Munich-based publishers.
Google’s AI had falsely tied the publishers to scams, subscription traps and dubious business practices, the court said, mixing them up with genuinely disreputable firms and inventing links that appeared in none of the cited sources. The injunction applies across the European Union, and the court said its reasoning could carry international weight. The decision is not final.
The ruling turns on a distinction. A traditional search engine merely points users to outside websites, which under German case law gives operators only limited, indirect liability. AI Overviews, the court found, go further by generating independent, new statements that rewrite and combine third-party content “in its own words and according to its own structure.” Because Google alone controls the system and its algorithms, the court treated the output as Google’s own speech and the company as a direct infringer. It also gave AI-generated claims weaker free speech protection, calling them the product of an algorithm rather than a held conviction, and an expression of Google’s business activity.
Google argued that users could check the linked sources themselves and generally knew not to trust AI blindly. The court rejected that, finding the overview was self-contained and understandable on its own, and drawing a parallel to press law, where publishers answer for standalone teasers even if readers never open the full article.
Research showing that users almost never click source links supported the point. Google could not fall back on Digital Services Act host protections or standard notice-and-takedown rules. The court ordered Google to cover 80% of the legal costs, with the plaintiffs paying 10% each.
Why the Industry Is Watching
The decision is one of the first to assign responsibility when a generative AI system gets facts wrong, and its answer is blunt: the company that built it. Google said it is “carefully reviewing this decision, which is not yet final,” and maintains that its AI Overviews are designed to reflect information already on the web and are accurate in the large majority of cases.
A lawyer for the publishers framed the outcome differently, saying providers “cannot hide themselves behind AI” when their systems misrepresent people. The principle, if upheld on appeal, would expose any operator whose AI paraphrases web content, including OpenAI’s ChatGPT, Anthropic’s Claude and Perplexity.
The Bigger Test
The scale of the problem extends well beyond one case. An analysis by the AI startup Oumi for the New York Times found that Google’s AI Overviews, running its Gemini 3 model, answered correctly about 91% of the time. That is reliable enough for everyday use, but at Google’s volume it still implies millions of wrong answers every hour.
The same analysis found that 56% of the correct answers could not be traced to the sources Google linked, meaning the system often produces claims with no verifiable origin. Legal scholars note a parallel in the United States, where Section 230 shields platforms for third-party content but may not cover material an AI itself creates. The Munich court targeted exactly that gap.
