German court finds Google liable for false AI answers

Google can be held directly liable for false answers generated by its in-house artificial intelligence (AI), a German court ruled on Friday, in a case that could have major implications for tech companies.

The 26th civil chamber at the Munich Regional Court, which specializes in press and defamation law, issued the potentially landmark ruling. The decision is not yet final, as Google can appeal.

Two Munich-based publishers sued the internet company, alleging that its "AI Overview" feature, which appears at the top of searches, falsely linked the companies to fraud schemes, dubious business practices and subscription traps.

The AI mixed information about other, genuinely questionable companies with the plaintiffs and invented connections that did not exist even in the sources Google attached.

At the heart of the legal dispute was the question of whether an AI overview should be treated in law the same way as conventional search results. 

Google argued that the company is not responsible for the data processing and does not adopt third-party content shown in the overview as its own.

The court firmly rejected that argument, judging that the AI summary was not a mere display or linking of search results, but content attributable to the search engine operator itself. 

Because the AI summarizes results in its own words, evaluates them and presents them in a structured way, Google thereby creates entirely independent, new statements that go beyond the bare links, the judges argued.

Accordingly, the court found that existing case law from the Federal Court of Justice, which protects search engine operators from direct liability for the mere listing of third-party content, does not apply.

Google's line of defence - that users could check the sources themselves via the links and were aware "that AI-generated information should not be trusted blindly" - was also rejected by the court. 

The chamber emphasized that the AI overview constituted "a self-contained statement with independently comprehensible content." 

There was no indication for the reader of any unreliability in the content, which meant that the mere possibility of further research did not free the company from liability for reputational damage, the judges argued.

The court ordered Google to cease spreading the false claims and ordered the company to pay 80% of the legal costs. 

A Google spokesman said: "We invest heavily in the quality of AI overviews to ensure that the vast majority of answers provide accurate information." The company would carefully review the decision, which is not yet final, he said.