German Court Rules Google Directly Liable for AI Overview False Information

A German court has ruled that Google is directly liable for incorrect information presented by its AI Overviews platform, according to a report by The Decoder. The country has laws in place that protect search engine operators from liability, but the court ruled that this doesn't apply to AI overviews. It has classified Google as a direct infringer because the AI Overview is its own content and not just a list of search results.

This all started when the company's AI overview algorithm spread false claims about two Munich-based publishers. The publishers were tied to scams, subscription traps and shady business practices via certain search queries. The court says the AI jumbled up information about totally separate companies, drawing connections that didn't appear in any linked sources and didn't actually exist.

The publishers sent Google a cease-and-desist letter, but they say it didn't respond appropriately. The Regional Court of Munich has hit Google with a temporary injunction in which it is no longer allowed to spread false information about the two companies involved in the case.

The ruling places the onus of responsibility for any factual errors on Google, as the AI Overview rewrites information "in its own words and according to its own structure." In this case, the overview confidently suggested that one of the publishers was "known for dubious business practices" and built its own structure with a summary, red flags for these shady practices and tips for users. The problem, again, is that AI Overview was actually pulling information about another entity. It even invented claims out of thin air that weren't noted in search results.

Simply put, Google the search engine wouldn't be directly liable because all it does is make third-party content findable. Google the AI Overview operator is liable because these overviews create "independent, new and substantive statements" by evaluating and combining content from third-party websites. This is pretty much what I do for work, and I certainly would be liable if I made up a bunch of slanderous stuff about a couple of publishing companies.

Google argued at the hearing that users could check the linked sources to verify if the AI summary was correct. It also said that these users knew "that information generated with AI should not be blindly trusted." That's a fairly remarkable statement given the speed in which these Gemini overviews were foisted upon us. Studies suggest that just one percent of users click on source links after reading one of these overviews.

Just how bad is the AI Overview factual accuracy problem? This depends on your perspective. A recent study reported on by The New York Times suggested that the overview gets stuff wrong around nine percent of the time. That sounds like a decent enough metric until you realize just how busy the platform is. Google recently crowed that 2 billion people interact with AI Overviews each month. That's 24 billion people each year. Some simple napkin math suggests that translates to well over 2 billion incorrect queries every year. This is an extremely conservative estimate, given how ubiquitous the platform has become and how many Google searches (16.5 billion) are performed each day.

However, factual accuracy isn't the only problem here. A study noted that these overviews also have a sourcing issue. This analysis found that 56 percent of correct answers couldn't actually be backed up by the linked source. In other words, there's no real way for users to actually check the AI's work.

As an aside, there is a way for users to "miss out" on all of this magical AI search technology. Just pop "-ai" at the beginning of any Google query. You're welcome.