Munich Regional Court I, judgment of 28 May 2026 – 26 O 869/26 (interim injunction proceedings)
The court prohibited Google from displaying infringing claims about two publishing companies in its AI-Overview. IT regarded the AI summary as a statement made by the search engine operator itself – for which Google is liable as the direct infringer(unmittelbarer Störer).
If one entered the name of a Munich-based publishing house into Google together with the (suggested) term scam („Betrugsmasche”), the AI Overview function generated a summary. This summary linked the publisher to dubious business practices, spoke of subscription traps and drew connections to other companies – allegations that could not be substantiated or actually concerned other firms.
The court upheld Google’s liability as a direct infringer. It regarded the AI overview as Google’s own content, not merely a display of third-party search results:
- The AI does not merely display results as links or previews, but summarises them in its own words and structure, evaluates them, and even confirms the query in its introduction (like “Yes, … is known for dubious business practices…”).
- In some cases, the overview contained statements that were not found in any of the linked sources.
- Since Google introduced and offers the AI itself – and only Google has influence over the service and algorithms – it must be held accountable for its results.
It is noteworthy that the court expressly rejects the liability exemptions otherwise customary for search engines:
- No search engine privilege: Unlike a simple list of results, new, distinct statements are generated here, which Google – by cross-referencing with the underlying sources – can also verify itself. Furthermore, the AI overview is not essential for using the internet – unlike search engines in general; the flood of data is already manageable through the display of links. Liability cannot be limited here to obvious infringements of the law. Otherwise, a gap in protection for the personal rights of those affected would arise.
- No application of autocomplete case law: The AI overview goes beyond terms displayed via autocomplete because it formulates its own response – rather than merely suggesting a search term.
- No exoneration through source links: The fact that users can click on and check the sources does not relieve liability. The overview is comprehensible in its own right; the mere possibility of refuting a statement through further research does not absolve one of responsibility for it.
- AI Regulation definition as an argument for attribution: Interestingly, the court refers to the provider definition in Article 3(3) of the AI Regulation when considering whether Google should be classified in charge for the AI output. This may be an indication that future AI Regulation obligations (e.g. transparency obligations, risk management, etc.) may be regarded as standards of care under civil law.
The judgment was delivered under the law of defamation and personality rights (Sections 1004, 823 of the German Civil Code (BGB) in conjunction with corporate personality rights) and follows a growing line of case law (we reported on the decision of the Higher Regional Court of Hamm, judgment of 12 May 2026 – I-4 UKl 3/25 on liability for an AI chatbot on a website).
Relevant for practice:
- Data protection law: The GDPR did not apply here because the data subjects were legal entities. If, on the other hand, an AI generates incorrect statements about natural persons, claims for rectification and erasure (Articles 16 and 17 of the GDPR) as well as responsibility for data accuracy (Article 5(1)(d) of the GDPR) may apply.
- Corporate and personal data protection: False or otherwise unlawful AI statements about companies or individuals may give rise to claims for injunctions and, where applicable, damages – as both a data subject and an operator, you should be aware of both perspectives.
In practice, when using AI functions, it is advisable to implement close monitoring of the output (e.g. linking to verified sources, topic filters and guardrails, escalation to human review, monitoring) as well as effective reporting and correction channels for data subjects.
This is a first-instance ruling in summary proceedings; a ruling by the highest court is still pending. However, a trend in case law towards affirming liability for AI content may already be discernible.
(Dr. Marian Klingebiel, Unverzagt Law)