Munich court holds Google responsible for errors in AI Overviews

A court in Munich, Germany, has issued a landmark ruling: Google must take responsibility for errors in its AI Overviews, the summaries that appear in search results and compile information using artificial intelligence.

The court made Google answerable for every fact

The Munich court ruled that an AI-generated search summary is not simply an automatic retelling of the articles it found. It is compiled content, and Google itself bears full responsibility for it. If an AI Overview contains an error or presents incorrect information, the company must correct it, add the right source or remove the material entirely.

The judge noted that when Google uses an algorithm to aggregate text, it becomes the publisher of the resulting content, not merely a technical channel for delivering information. The distinction is fundamental. A newspaper is responsible for every fact it publishes. Google is now held to the same standard.

Google's response and position

Google objected that in the vast majority of cases the overviews are accurate and useful to users. The company stressed that it is investing aggressively in improving the quality of AI Overviews and is continually working to reduce errors. Here is what Google says it has already put in place:

• Multi-level checks before each overview is published
• Intelligent filters to exclude unreliable sources
• A user feedback system for quickly identifying errors
• Regular model updates based on problems that have been found
• Labels showing the model's confidence level on contested topics

Google also stressed that this is the first court ruling of its kind in the world and that its impact so far remains local, within German jurisdiction.

A precedent for AI search

The Munich ruling may become a precedent for other European jurisdictions and even for the US. If several major economies adopt similar rulings, it will put additional pressure on companies that use generative AI in critical functions.

The core problem is that AI hallucinations are a systemic risk, not a rare bug. Language models can invent facts, cite nonexistent studies and mix up dates, names and events. When such content is presented in search as objective fact, it can mislead tens of millions of users.

"When

Google chooses which text to compile and how to rework it, that is still Google's choice, and therefore its responsibility" — such was the court's logic.

What this means

The ruling shows that regulation of AI content starts not with bans but with transparency and accountability. Google has not been forbidden from using AI for search, but it is obliged to guarantee the quality and accuracy of results.

This will inevitably raise development costs: companies will need qualified teams of reviewers, a multi-level verification pipeline and a fast response to complaints. For users, that is good news: they will get more reliable search overviews. For innovation in AI search, it is a challenge, because the level of required accountability has risen sharply.