In a landmark decision that could reshape the tech landscape, the Regional Court of Munich has issued a temporary injunction against Google, stripping away the legal shield that historically protected search engines from liability. For years, companies like Google were viewed as mere conduits—digital librarians pointing users toward third-party content without being responsible for the accuracy of what those sites said. However, the court ruled that Google’s AI Overviews are a fundamentally different animal. Because the AI synthesize, rewrites, and structures information into a “new” statement rather than just presenting a list of links, the court classified Google as an editor—a direct infringer—rather than just a search facilitator. This shift means that Google can no longer hide behind traditional search provider protections, as the AI’s output is now viewed as the company’s own proprietary content.
The case emerged after two Munich-based publishers found themselves falsely maligned by Google’s AI. When users searched for these companies, the AI confidently generated summaries—often starting with phrases like “Yes, [company] is known for dubious business practices”—that falsely linked the plaintiffs to scams, subscription traps, and shady operations. The court was particularly struck by the fact that these damaging claims did not exist anywhere in the sources the AI actually cited. By hallucinating connections that the cited websites never made, the AI created its own truth. The court dismissed Google’s defense that the AI’s behavior was merely a reflection of web content, noting that Google has total control over the algorithms driving these summaries. Essentially, if you build the brain and set the rules for how it processes information, you must shoulder the blame for the misinformation it produces.
Google’s defense team attempted to argue that users are responsible for verifying facts and that the public understands AI shouldn’t be blindly trusted. The Munich court was having none of it. They pointed out that, in practice, users rarely click on the source links provided in an AI overview, preferring the convenience of the AI’s “self-contained” summary. By presenting these answers as authoritative, factual bulletins, Google created an environment where users are conditioned to stop digging deeper. The court drew a sharp parallel to traditional press law: if a newspaper prints a false headline that harms a reputation, they are liable, even if the reader doesn’t finish the article. Google’s argument—that we should assume AI is unreliable—was viewed by the judges as a self-defeating admission that undermines the very utility of the feature they are pushing onto billions of people.
Furthermore, the ruling struck a blow to the idea that AI-generated text enjoys the same free speech protections as human-written journalism. The court argued that AI output is the result of an algorithm processing data, not an expression of an individual’s personal conviction or intellectual labor. Because Google’s AI service is primarily a business function, it does not hold the same protected status as a citizen expressing an opinion. Consequently, when the AI accidentally defames a company by linking them to scammers, the privacy and reputational rights of the victim carry more legal weight than Google’s business interest in maintaining a “convenient” search feature. This effectively forces Google to ensure its AI is legally sound before it hits the “generate” button.
The implications of this ruling extend far beyond Munich, posing a massive scalability problem for the entire AI industry. While studies suggest that models like Gemini might be accurate 91% of the time, the sheer volume of global searches means that the remaining 9% translates into millions of instances of misinformation every single day. If legal systems worldwide adopt the Munich court’s reasoning, the cost of being “mostly accurate” becomes prohibitively high. AI companies are currently treating these tools as helpful assistants, but the court has signaled they are legally functioning as content publishers. When you account for the fact that a majority of AI answers cannot even be backed up by the sources they cite, the potential for widespread defamation lawsuits becomes a existential threat to companies like Google, OpenAI, and Perplexity.
Ultimately, this case serves as a warning shot that the “Wild West” era of generative AI is coming to an end. By rejecting the “notice-and-take-down” protections that previously kept Google safe, the court has effectively put the burden of truth on the creators of the software. Google has been ordered to pay the vast majority of the legal fees, and unless they can figure out how to stop their algorithms from fabricating defamatory accusations, they face a future of endless litigation. Whether on appeal or as a precedent for future international rulings, this decision highlights a simple, uncomfortable truth: if a tech company wants the credit for building a “smarter” search engine, they have to accept the liability for the lies their machine tells along the way.