A German court held Google directly liable for false claims generated by its AI Overviews

In a ruling dated May 28, 2026 and made public in June, the Regional Court of Munich I (LG Munich I) prohibited Google from disseminating untrue factual claims about two Munich publishers through its AI Overviews, classifying Google not as a neutral intermediary but as a direct disturber responsible for its AI's output. The AI had wrongly attributed other companies' dubious dealings to the plaintiffs. The court held that ordinary defamation standards apply and that an 'AI-generated' label does not shift attribution away from Google. Google said it would appeal.

Google · Incident May 28, 2026 · Indexed Jul 10, 2026 · 2 sources

Records by entity: Google

Online providers cannot hide behind the fact that a statement was generated by AI; they are liable for its output and its publication.
What
In a ruling dated May 28, 2026 and made public in June, the Regional Court of Munich I (LG Munich I) prohibited Google from disseminating untrue factual claims about two Munich publishers through its AI Overviews, classifying Google not as a neutral intermediary but as a direct disturber responsible for its AI's output.
Incident date
May 28, 2026
Who
Google
Failure mode
Hallucination
AI surface
Search / RAG
Severity
High

What happened

Two Munich-based publishers sued after Google's AI Overviews attributed the dubious dealings of other companies to them, synthesizing claims that were not supported by the underlying search results. In summary proceedings, the LG Munich I prohibited Google from disseminating the false statements and classified Google as a direct disturber whose AI produces false information as its own content, rather than as a search engine merely surfacing third-party material. The court held that because the AI independently analyzes, weights, and synthesizes content into a new statement, the limited liability that shields search snippets does not apply, and the "created with AI" label does not change attribution. Google must remove the content, prevent comparable false claims, and bear roughly 80 percent of costs; it said it disagreed and would appeal.

What broke inside the model

Failure path · mode profile · Hallucination
  1. 01 · TriggerA user asks for a fact, a citation, or a figure.
  2. 02 · Model stepThe model writes a fluent, confident answer.
  3. 03 · Control gapNothing ties the claim back to a real source.
  4. 04 · FailureA fabricated fact ships as if it were verified.
  5. 05 · ConsequenceThe false claim reaches a customer, a court, or the public.

Confidence holds, and even spikes, as the claim detaches from any source.

AI Overviews generated a defamatory factual claim that the cited sources did not support, a hallucination that fused unrelated companies' misconduct onto the plaintiffs. Because the feature summarizes and evaluates results in its own words rather than quoting them, the court treated the output as Google's own statement. The system produced a confident synthesis with no grounding in the retrieved material, and published it at scale.

Public visibilityHigh
Regulatory exposureActive
Customer impactFew customers
Financial impactDisclosed
Time to disclosureMonths
  1. PressLG Munich I: Google ordered to pay for false statements in AI summariesheise.de
  2. PressGoogle Handed AI Liability Blow in German Ruling That Could Transform AI Searchlaw.com
Permalinkhttps://failureindex.ai/failures/lg-munich-google-ai-overviews-direct-liability
CitationAI Failure Index. "A German court held Google directly liable for false claims generated by its AI Overviews" (FI-0716). Realm Labs. https://failureindex.ai/failures/lg-munich-google-ai-overviews-direct-liability (indexed Jul 10, 2026).
Share cardA branded image of this record for posts and slides.

Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0716. Full dataset at /data.

Note from Realm Labs, the Index steward

How Realm would have caught this

Controls for this failure mode
  • Prism
  • OmniGuard
  • AI Detection & Response (AIDR)

Realm checks a synthesized claim against the sources it purports to summarize and flags output that asserts facts the retrieved material does not support. For a generative search surface, that grounding check at runtime is what keeps an unsupported, reputation-damaging synthesis from being published as the provider's own words.