By Hamza Faisal
A German court has ordered Google to stop its AI from speaking for it. On September 24, the Aachen Regional Court issued an emergency injunction barring Google Ireland Limited from spreading three text passages its AI Mode generated about a physician — passages that pinned another man's criminal past on the doctor.
According to reporting on the ruling, the AI Mode responses claimed the former chief physician had harassed his secretary, been dismissed, received a suspended sentence for sexual assault, and walked away with a six-figure severance package. None of it was true. The AI had apparently taken the story of a different former senior employee and attached it to the doctor's name.
The doctor had complained in detail by email in late August and asked Google to remove the false claims. In mid-September the company told him it would take no action; a formal legal demand from his lawyers changed nothing either. The court noted the false texts were still live.
A €250,000 warning shot
Because the fabricated allegations could seriously damage the doctor's career, the court acted on an emergency basis — without hearing Google first. Each violation of the order can cost up to €250,000, per reporting on the case.
But the fines are almost beside the point. The core of the ruling is how the court classified the AI's output. The judges treated Google's AI answers not as neutral links to third-party websites but as Google's own statements — "independent, new, and substantive statements based on an evaluation and combination of content from various third-party websites."
"The respondent provides AI-generated search and is therefore also responsible for the answers generated in this context," the court wrote.
Google could have checked — its own AI would have caught it
The court also rejected the idea that these mistakes were unavoidable. Google could verify its AI's claims, the judges said — for example, by comparing the generated text against the source pages it was drawn from.
In a pointed observation, the court added that if Google had confronted its own AI with the doctor's objections, the system would have caught the errors itself. The capability to self-correct existed; it just wasn't used.
The Munich ruling that wouldn't go away
Aachen is not a first. In late May, the Munich Regional Court I reached the same conclusion in a case where Google's AI Overviews falsely linked two Munich-based publishers to fraud schemes. The Aachen court cites the Munich ruling directly — a sign that a consistent legal view is forming across Germany's courts.
Google announced an appeal in the Munich case, then quietly settled before any higher court could rule on the reasoning. The settlement terms were never disclosed. The Aachen decision shows the legal argument is only getting stronger in the meantime.
And the courts are not the only pressure. In July, Germany's ZAK media regulator ruled that Google's AI Overviews and Perplexity fall under national media law — AI answers are providers' own content, not shielded intermediary hosting. Google is appealing that ruling too.
If the view holds that AI answers are Google's own words, the implications go far beyond one doctor or two publishers. Publishers could demand payment for the material the AI draws on, and the entire operating model of AI search in Germany would have to be rebuilt. The injunction is preliminary — Google can still challenge it — but the direction of travel is now unmistakable: when your AI speaks, you answer for it.
