Google Is Liable for Its AI —
And You're Liable
for Yours

In a preliminary injunction (case no. 26 O 869/26), Munich Regional Court I made it clear: anyone who publishes AI-generated content as their own statement is liable for it — without limitation. For mid-sized businesses, that means every AI output needs a quality loop before it goes public.

What exactly did Munich Regional Court I decide?

Two Munich publishing companies sued Google, represented by Lausen Rechtsanwälte. The reason: Google's AI Overviews had characterized the publishers as "known for dubious business practices" and linked them to fraud schemes and debt-collection claims.

The court found: these connections "did not appear anywhere in the linked sources at all." The AI had constructed them on its own.

The decisive point: the court classified Google as a direct interferer. An "AI overview" is "fundamentally different from a normal results list" — it's Google's own statement, not someone else's content. The classic host-provider privilege no longer applies. Google bears around 80% of the litigation costs.

Why does this affect every mid-sized business?

The ruling isn't an isolated case. Back in September 2025, Frankfurt Regional Court ruled similarly on Google's liability. The legal line is solidifying. According to the court, the Munich ruling could have effect not just in Germany, but EU-wide.

Specifically, the following legal grounds apply to SMBs for unreviewed AI output:

• Section 5 UWG (misleading practices): when AI-generated statements contain false information about products or competitors.
• Sections 823, 1004 BGB (personality rights): exactly the legal basis of the Munich ruling — applicable to any company that publishes AI statements about third parties.
• EU AI Act, Art. 13 and 14: for high-risk use, transparency and oversight obligations apply — including human-in-the-loop.

The excuse "the AI said it" is legally worthless after Munich.

What do safe AI workflows look like?

AI without oversight is a legal risk. Safe AI workflows rely on three principles:

1. Human-in-the-loop as the standard. No AI-generated text leaves the company without human sign-off. That applies equally to website copy, proposals, emails, and social media posts.

2. Quality loops instead of blind trust. A QA system checks every output for facts, tone, and compliance — before a phone agent conducts a call or a text is published.

3. Source transparency. Every statement based on AI output needs a traceable source.

At Amazing Agents, human-in-the-loop isn't an optional add-on. Our phone agent Ben conducts qualifying calls, documents every contact, and passes the result along in structured form to the human team member — who then decides. That's not a detour. It's the only legally sound architecture.

What concrete steps should SMBs take now?

1. Take stock: Where do you use AI-generated text in external communication?
2. Assess the risk: Which of these texts could contain statements about third parties?
3. Set up a quality loop: Route every AI output through a review process.
4. Document it: Keep a traceable record of which texts are AI-generated.
5. Get legal advice: Have a lawyer review specific individual cases.

The Munich ruling is no reason to panic — but it's a clear signal: AI output is business communication. With all the consequences that come with it.

FAQ — Munich Regional Court I and AI liability for SMBs

Is my company liable for AI-generated text on our website?

Yes. Under the Munich ruling (case no. 26 O 869/26), AI-generated content that a company publishes counts as its own statement — with full liability under Sections 823 and 1004 of the German Civil Code (BGB).

Does the ruling also apply to ChatGPT, Claude, or other AI tools?

The ruling concerns Google AI Overviews specifically. But the underlying doctrine applies to any AI tool that publishes content as its own statement.

What is a quality loop?

A review process in which every AI-generated output is checked for accuracy, tone, and compliance before publication. Human-in-the-loop is the structural answer to the liability risk.

Is an "AI-generated" disclaimer enough protection?

No. Labeling content doesn't protect against liability. What matters is whether the statement is accurate — not whether it's marked as AI-generated.

Which industries are especially affected?

Financial services, healthcare, and any company with AI-generated customer touchpoints carry a particular liability risk. But in principle, the ruling applies across all industries.

Note: This article does not constitute legal advice.