German court rules retailers liable for AI chatbot errors

A German appeals court has ruled that online retailers are fully liable for misleading information generated by their AI chatbots. Yvonne Bachmann reports for Händlerbund News that the Higher Regional Court of Hamm (Oberlandesgericht Hamm) issued the ruling on 12 May 2026 (case reference: 4 UKl 3/25).

The case was brought by the consumer protection organisation Verbraucherzentrale North Rhine-Westphalia against a medical services company. The company’s chatbot had independently invented false specialist medical titles for its managing directors when responding to customer enquiries. The company argued the AI system acted largely autonomously and that its errors were unforeseeable.

The court rejected this argument. In its reasoning, it stated that using technical tools such as machine learning algorithms counts as normal conduct by the operator. Even if an AI language model generates responses without human involvement, this does not release the operator from responsibility. The court also clarified that liability applies even when a system was originally trained on entirely correct data.

Prevention, not excuses

A key factor in the ruling was that the company, after receiving a cease-and-desist letter, successfully blocked the false outputs using keyword filters and specific prompt instructions. The court concluded that what proved technically feasible afterwards should have been implemented before the chatbot went live.

The judges also dismissed the notion that customers are naturally sceptical of AI-generated information. On the contrary, the court noted that many users place particular trust in machine-generated data, as they tend to perceive computers as less error-prone than humans.

Because the case raises fundamental questions about legal attribution in AI systems, the court has allowed an appeal to the Federal Court of Justice (Bundesgerichtshof). The ruling is therefore not yet final.