Berlin Court Rules Google AI Overviews Are Just New Search Formats

A recent ruling from a Berlin court has created a significant legal pivot point for generative search by categorizing Google's AI Overviews as a new display format rather than original content. This decision contrasts sharply with recent judicial findings in Munich, leaving the tech industry grappling with the looming question of liability in the age of AI-driven information retrieval.

The Berlin Verdict: Aggregation Over Authorship

In an early June ruling, a Berlin court concluded that Google’s AI-generated summaries do not constitute independent statements made by the search engine. Instead, the court viewed these summaries as a "new search result format" that merely aggregates and pulls together information from existing third-party websites.

The case originated from a lawsuit filed by a perfume company regarding trademark infringement. When users searched for fragrance imitations, the AI surfaced brand names alongside links to websites selling cheaper alternatives. The court ruled that this did not violate trademark or competition laws, arguing that the search engine was simply surfacing information already available on the web and lacked "decisive influence" over the specific phrasing of the answers. The court further posited that an average user would recognize the AI is simply summarizing external sources.

The Munich Contradiction: Direct Liability for Hallucinations

This decision stands in stark opposition to a Munich court ruling delivered just days prior. In the Munich case, Google's AI falsely linked two publishers to fraudulent schemes—claims that did not exist in any of the cited source material.

The Munich court rejected the notion that search engines are mere aggregators when the AI "invents" connections. It ruled that because Google controls the underlying AI models, system parameters, and response structures, it is directly liable for false factual claims. Crucially, the Munich court dismissed the argument that users bear the responsibility to fact-check, labeling AI summaries as independent content that requires accountability from the provider.

La tension entre ces deux décisions met en lumière une lacune réglementaire massive. L'affaire de Berlin portait sur le droit des marques et de la concurrence, où le résumé de l'IA a été considéré comme un reflet fidèle du contenu de tiers. L'affaire de Munich portait sur l'exactitude factuelle et la diffamation, où l'« hallucination » de l'IA a généré de nouvelles informations erronées.

Pour les fondateurs et les développeurs d'outils de recherche intégrant des LLM, les implications sont profondes. Si la logique de Berlin prévaut, les entreprises pourraient bénéficier d'une immunité juridique significative en affirmant que leur IA n'est qu'un outil d'indexation sophistiqué. Cependant, si la logique de Munich devient la norme, le coût du maintien de l'exactitude de l'IA deviendra une responsabilité juridique massive, car les fournisseurs seront tenus responsables de chaque affirmation « indépendante » générée par leurs modèles.

Points clés