German court bars Google's AI from falsely accusing a doctor of sexual assault

On August 28, 2026, a colleague warns a German doctor. When Google's "AI Mode" is asked why the doctor left their post, the answer mentions a sexual assault against their secretary. Follow-up searches, with words like "suspension" or "legal proceedings," add details, even invented court consequences. All of it is false: these facts concern another former university professor.

On September 24, the Aachen regional court (Landgericht) issued a preliminary injunction. It bars Google from publishing three versions of the contested texts, under penalty of fines of up to €250,000 per violation, according to the specialist site The Decoder. The decision, published as early as October 8 by the legal site beck-aktuell and reported in particular by the daily Frankfurter Allgemeine Zeitung and by Next on October 20, is not final: Google can challenge it.

Warned in late August, Google chose to do nothing

The doctor was head of department and professor of neuroradiology from 2010 to the end of June 2026. In late August, they wrote to Google to report the error and request removal. In mid-September, the company replied that it would do nothing, according to The Decoder. A formal notice sent the same day had no effect, and the texts stayed online. According to Next, the AI claimed the doctor had been suspended in February 2024 after a complaint from their secretary, then convicted in March 2026.

The court ruled without first hearing Google, because of the risk of serious harm to the doctor's reputation. Its reasoning comes down to one idea: Google offers an AI-generated search, so it is answerable for what that search produces. In the judge's view, these answers are new, autonomous statements built from third-party content, not a mere link to pages. The judge adds that Google could have checked the texts against their sources, and that by confronting its AI with the doctor's objections, it would probably have discovered the errors.

A second ruling in four months

The Aachen court explicitly relies on a precedent. In late May 2026, the Munich I regional court had already held Google directly liable for false statements in its "AI Overviews," the generated summaries at the top of results, which linked two Munich publishers to scams. Google announced an appeal, then reached a settlement whose terms are unknown. No higher court has therefore ruled.

The Decoder sees a perverse effect. By settling, Google avoids a high-level precedent, but first-instance courts keep applying the same reasoning. The site even imagines specialized law firms starting to hunt for defamatory answers in order to obtain injunctions and settlements.

What this changes for companies

For a person or company targeted by a false answer from an AI search tool, these rulings open a path to action, at least in Germany. The course of this case offers a few markers:

• keep dated screenshots of the answers, along with the queries that produce them;
• report the error in writing to the provider and keep a record of its reply;
• if it refuses, an emergency procedure succeeded here in less than a month.

For companies that deploy a public chatbot, the lesson is the reverse: a judge may treat what the tool says as their own statements. The real question, which no appeals court has yet settled, remains open.