Two German courts have recently decided that companies must take responsibility for the errors of their AI systems. The Munich District Court ordered Google to stop its AI summary falsely accusing a publishing house of fraudulent practices. The Hamm Higher Regional Court had previously held a clinic liable for a lying chatbot.
Munich Ruling Binds Google to Its Own AI Statements
The trigger for the Munich proceedings was the “AI Overview” feature in Google Search. When users searched for a Munich publishing house, the AI displayed a summary attributing dubious business practices, subscription traps, and collection demands to the company. These allegations did not appear in the linked sources; according to LTO, the AI had either fabricated them or confused the company with others. The publisher refuted the claims with its own evidence and took the matter to court. On May 28, 2026, the Munich District Court ruled under case number 26 O 869/26 that Google could not treat the statement as a neutral reproduction of third-party search results. The chamber classified the AI response as Google’s own statement, attributable to the corporation, and weighed the publisher’s corporate personality rights against freedom of expression. An assessment resting on demonstrably false facts significantly restricts freedom of expression, the chamber found in substance. Google must now refrain from repeating the statement and risks a court-imposed fine if it does so again. An appeal to the Munich Higher Regional Court remains legally possible, though no such step has been made public so far.
Hamm Ruling Holds Chatbot Operators Accountable
Two weeks earlier, the Hamm Higher Regional Court had decided a similar case. The chatbot of a cosmetic clinic had attributed fictitious specialist titles, such as “specialist in aesthetic medicine,” to the clinic’s managing directors, a qualification they did not actually hold. The Consumer Center North Rhine-Westphalia, a consumer-protection body, had filed the lawsuit. The court ruled on May 12, 2026, under case number 4 UKl 3/25, that the false statements constituted a business act within the meaning of Germany’s law against unfair competition. The clinic could not claim the chatbot had acted independently and unpredictably. It communicates on the company’s behalf and therefore remains its tool, not an independent third party. The court ordered the practice to stop using the misleading titles and to bear the legal costs. Consumer advocates view the ruling as a signal to the entire health and cosmetics industry, where chatbots increasingly replace initial phone consultations. Both chambers arrive at a consistent principle: whoever deploys AI systems in customer contact bears the risk of erroneous output, regardless of whether they built the model themselves or merely integrated someone else’s technology.
Further Rulings Intensify the Trend Toward AI Liability
According to an analysis by Datenschutzticker from July 9, 2026, the two rulings join a growing number of decisions on AI liability. Courts outside Germany are following a similar path: in July 2026, India’s Supreme Court struck down lower-court decisions built on fabricated AI-generated case citations and demanded, according to Republic World, “zero tolerance” for such hallucinations. The court classified the use of fabricated precedents as misconduct by the lawyers who submitted them, not as a mere mistake, and asked the bar association to set up its own review panel for AI use. In the United States, comparable cases are also multiplying, as courts work out how far companies’ responsibility for their AI tools extends. A common pattern is emerging internationally: courts increasingly ask who controls and deploys an AI system, rather than how mature the underlying technology is. For German companies, that likely means reviewing existing chatbot and AI features early, before further proceedings bring added costs and reputational damage.
It remains open whether either proceeding will reach Germany’s Federal Court of Justice and set a nationwide standard for AI attribution. Until then, companies must weigh the risk case by case whenever they delegate customer contact to language models.
