A Court Just Ruled That Your AI's Answers Are Your Words. The Liability Is Now Yours.

What the Munich court actually decided

For years the assumption was that an AI assistant simply surfaced what was already on the web, and that responsibility stayed with the original source. The Regional Court of Munich rejected that. It ruled that Google's AI Overviews generate independent, new, substantive statements by evaluating and combining third-party content, and that only Google can check those statements before they appear. The court issued a preliminary injunction barring Google from repeating false claims about two publishing houses, under threat of a fine of up to 250,000 euros.

The distinction matters more than the parties involved. A list of links with quotes points elsewhere. An AI answer speaks in its own voice. The moment a system composes a fresh claim, the court treated the operator as the author of that claim. That is the line most businesses deploying AI now sit on the wrong side of by default.

Why this reaches far beyond Google

The ruling is one of the first in Europe to hold an AI operator directly liable for AI-generated content, and the logic does not depend on Google. A support chatbot that calls a competitor fraudulent, a sales assistant that invents a product claim, an internal tool that states something false about a supplier: each produces a new statement that only the operator can control. Under the Munich reasoning, the operator answers for it.

Google argued that anyone could disprove the statement with further research. The court found that the ability to refute a claim does not normally remove liability for making it. For any company that shipped an AI assistant on the assumption that the model, not the business, carries the risk, that assumption is now a documented exposure.

What a serious company should do now

Treat every customer-facing AI output as published content, because a European court now does. That means knowing exactly where your AI generates free-form claims about people, companies, or facts, what guardrails sit in front of those outputs, and who is accountable when one is wrong. Most organizations cannot answer those three questions today, which is the real gap the ruling exposes.

The fix is not to switch the assistant off. It is to map the exposure: inventory the systems that speak in your name, constrain them where the cost of a wrong answer is high, and document the controls so the liability is managed rather than discovered in court. Done early this is a governance task. Done late it is a 250,000 euro lesson.

Read next: Your AI Agent Can Now Spend Your Money. The Hard Part Is Everything That Happens After.  ·  There Is Now One Shared Reference for AI Safety. Here Is Why It Reaches Your Procurement.