Das Google gets walloped

A German regional court has issued a preliminary ruling that Google is directly responsible for false claims on its AI-generated search results – the ones that now routinely appear as summaries at the top of many search pages, which, incidentally, also cause news websites to lose click-through traffic.

The Munich Regional Court has placed a temporary injunction on Google, preventing it from spreading false claims about two medium-sized Munich-based publishers through its AI-generated search overviews. If you had searched for the companies and taken the advice of Google’s autocomplete feature to add the word ‘scam’, what the summary turned up was not a set of links to or about the companies, but an AI-generated summary erroneously linking the publishers to scams, subscription traps and fraudulent schemes.

The AI results had also linked the publishing companies to shady companies, inventing connections that didn’t exist. The Google AI summary cited one source as a law firm issuing a warning notice about the publishers. But the warning notice didn’t actually relate to the publishers at all but to another company which is in fact suspected of engaging in illegal practices. The aggrieved publishing companies, which between them own twelve publishing brands focused on – believe it or not – technology and history, argued the Google AI summary had nothing to do with them. The AI was plain wrong.

For its part, Google relied on existing German case law, enshrined in the EU’s Digital Services Act, which protects search engine platforms from direct liability when they publish listings of third-party content. The platform, when it acts as a passive conduit for third-party information, is protected from liability unless they have been notified of an obvious legal violation and failed to act.

Google argued that it was not responsible for the data processing itself and that it did not adopt the third-party content featured in the overview as its own. The Munich Regional Court disagreed.

The court said Google’s AI summaries don’t just show or link to search results but are instead distinct pieces of content created by the search engine operator. The AI summarises results in its own words, having evaluated the content, and presents those summaries in a structured format creating an entirely new, independent statement that goes beyond merely linking to sites. In other words, there’s a big difference between search engine results and AI-generated summaries. And it can’t be our responsibility as users of search engines to know we shouldn’t necessarily trust what these-AI generated summaries tell us.

The finding, which Google says it will appeal, cracks open the search giant’s reliance on the argument that when AI hallucinates, it’s just some technical bug, for which the engine itself cannot be held responsible. Rather, the court held that this is content produced and disseminated by Google. If the ruling holds, then responsibility for the harm falls on the provider whose AI functionality got it wrong.

This must sound like a victory too, albeit in a faraway jurisdiction, for the Sydney-based graphic designer working for the Sydney Morning Herald who in 2025 created an illustration about an historical child murder case in which the alleged murderer was named under parliamentary privilege. The graphic designer was, in usual newspaper practice, named as the illustrator in the published story. Google’s AI summaries picked up the illustrator’s name and that of the victim – three-year-old Cheryl Grimmer who had gone missing in 1970 – and spat out a damaging and very wrong assertion that the illustrator was the alleged murderer. Google quickly corrected the error, but it’s hard to imagine the impact of being wrongly associated with such a heinous crime.