Landmark Ruling Says Google Liable for False Information Generated by Its AI

At Google’s Mountain View campus in California, Google announced one of the biggest changes in the history of its search engine at the main stage of the Google I/O developers conference in 2024. The company introduced AI Overviews for the first time, a system that in many cases replaces the traditional list of links with an AI-written summary displayed at the top of the page. For Google, it marked the beginning of a new era in which the search engine not only points users to information across the web, but also reads it, processes it and delivers a ready-made answer. The move signaled the official arrival of artificial intelligence at the core of the company’s most important product, and changed the way billions of people consume information every day. But that technological vision has recently suffered a major legal blow that could change the rules of the game for the company entirely.

A court in Germany ruled that Google will be directly responsible for false information that appears on its AI summary platform. The country, which has clear laws protecting search engine operators from liability for third-party content, made clear through the court ruling that those protections simply do not apply when it comes to AI summaries. The judges classified Google as a direct infringer because the system rewrites the information in its own words and according to an internal structure it determines, thereby generating entirely independent text rather than merely presenting a list of results.

The legal case began after a report by The Decoder, according to which Google’s AI algorithm spread serious false claims against two Munich publishers. The system linked the entities to fraud, “subscription traps” and questionable business practices through specific search queries. Examination of the case found that the AI mixed up information about entirely different companies and created connections that did not appear in any of the linked sources and did not exist in reality. The summary stated with confidence that one of the entities was known for problematic conduct, while presenting summaries and tips for users, when in fact the information was drawn from a different business entity and claims were invented from nothing that had not appeared in the original search results.

The publishers sent Google a warning letter, but after receiving no adequate response, they turned to the courts. The Munich Regional Court accepted their request and issued a temporary injunction prohibiting Google from continuing to distribute the false information about those companies. During the hearing, Google tried to defend itself and argued that users can check the attached links to verify the accuracy of the summary. The company also claimed that users understand that information produced by artificial intelligence should not be trusted blindly. That is a particularly problematic argument, given the speed with which these summaries are imposed on users. Recent studies show that the reality on the ground is entirely different, and only 1 percent of users actually bother to click the links after reading the summary.

The scale of the inaccuracy problem in AI results is proving to be broad and troubling. A study published in The New York Times shows that the platform provides false information in 9 percent of cases. Given the company’s reports of two billion monthly users interacting with the system, about 24 billion a year, a quick calculation shows that this amounts to more than two billion false queries every year. This is a very conservative estimate, given that about 16.5 billion searches are carried out on Google every day. In addition, another study found a serious source problem, in 56 percent of answers that were found to be correct, the information was not supported by the attached link at all, preventing users from scrutinizing the algorithm’s work.

For anyone who wants to avoid this technology, there is a simple way, adding the term “-ai” at the start of any search query will return you to the traditional and familiar results.