German court holds Google liable for AI Overview claims
TL;DR:
- The Regional Court of Munich found that Google’s AI Overviews are the company’s own content, not protected like ordinary search results.
- A temporary injunction bars Google from repeating false AI-generated claims that wrongly linked two publishers to scams and subscription traps.
- The ruling opens a route for businesses to challenge fabricated AI summaries as Google’s own statements — a precedent UK firms deploying generative AI should watch.
A Munich court has ruled that Google can be directly liable for false statements produced by its AI Overviews, rejecting the argument that the feature should enjoy the same liability protection as traditional search. Because Google’s AI rewrites, combines and evaluates information “in its own words and according to its own structure,” the court treated the output as Google’s own content rather than a neutral pointer to third-party pages.
Why the shield did not apply
The case involved two Munich publishers whose businesses were falsely tied to scams and questionable practices by AI summaries — claims that, the court found, did not appear in the linked sources. Google argued that established case law limiting liability for search engines and autocomplete should cover the feature. The court disagreed: unlike a list of links, an AI Overview generates new substantive claims and presents them as a complete answer, so users cannot reasonably be expected to cross-check the citations.
The injunction bars Google from repeating most of the disputed claims, and the company must pay 80% of the legal costs. The court noted the risk of repeat breaches because Google offered no cease-and-desist declaration with a penalty clause and the same algorithms could regenerate similar statements.
The logic rhymes with a question already live in the UK, where the NHS has been warned that doctors could be sued for AI mistakes: once an AI system authors a claim rather than merely surfacing one, who carries the legal risk?
Looking forward
The ruling is temporary and can be appealed, so it sets direction rather than settled law. But for UK businesses building products on generative AI, the principle is the warning: branding an output “AI-generated” does not outsource responsibility for it. As regulators and courts increasingly treat model output as first-party speech, “the model said it” is unlikely to survive as a defence.