A German court says you own what your chatbot says: OLG Hamm holds a clinic liable when its AI invented specialist titles

The problem

Aesthetify GmbH, a German provider of minimally invasive beauty treatments, put an AI chatbot on its website to answer visitors’ questions in real time and let them book appointments (source). The company’s two managing directors work at the clinic as doctors but hold no recognised specialist qualification in plastic and aesthetic surgery (source). On 3 April 2025, asked “Are A and B specialists in plastic and aesthetic surgery?”, the chatbot answered: “Yes, A and B are specialists in plastic and aesthetic surgery. They have the necessary expertise and experience to offer and perform individual aesthetic treatments. Would you like to make an appointment?” (source). In follow-up answers it called the two men “specialists in aesthetic medicine” and “specialists in aesthetic treatments” — and two of those three titles do not exist as recognised German qualifications at all (source).

What was built

No performance win is claimed here: this is an AI-governance liability case about a deployed customer-service chatbot, not a productivity outcome. The consumer association Verbraucherzentrale Nordrhein-Westfalen, a body entered on the list of qualified claimants under the German Injunctions Act, sued the operator for an injunction and reimbursement of its warning costs (source). After the warning letter the company deactivated and reprogrammed the bot, adding a prompt instruction to answer neutrally to any question containing the word “specialist” and a downstream keyword filter, and it admitted to the claimant that the answers had been a “regrettable technical error” (source). It refused, however, to sign a binding cease-and-desist declaration or to pay the warning costs, so the case went to court (source). In its defence the company argued that the chatbot ran autonomously on statistical probabilities, was not steered answer by answer, and that its outputs were therefore not attributable to the company as its own conduct (source).

The outcome

On 12 May 2026 the 4th Civil Senate of the Higher Regional Court of Hamm (Az. 4 UKl 3/25) held that the claim succeeded in full (source). The court ruled that the chatbot’s answers were the company’s own commercial acts within the meaning of the Unfair Competition Act, not the conduct of an independent third party (source); as the reporting legal press put it, “the chatbot is not a third party within the meaning of the law” (source). The decisive factor was control: the company set the chatbot’s operating framework and had enough influence over the system, shown by how easily it later reprogrammed the bot to stop making the false claims (source). The court also rejected the argument that consumers discount chatbot answers as error-prone, finding instead that many users trust computer-generated answers precisely because machines are perceived as less fallible than people (source). It ordered the company to cease and desist under threat of an administrative fine of up to 250,000 euros for each violation (source) and to pay 260.00 euros in flat warning costs plus interest (source). The judgment is not yet final: because the attribution of AI-generated statements raises new legal questions, the court admitted an appeal (Revision) to the Federal Court of Justice (source).

The weakest load-bearing link in this file is not a figure but a name: the published judgment anonymises the defendant as “the defendant” and its directors as “Dr A” and “Dr B”, so the clinic’s identity as Aesthetify GmbH of Recklinghausen rests on the secondary reporting (Legal Tribune Online, carrying the dpa wire) rather than on the primary document itself (source). Every quoted figure and holding, by contrast, is taken verbatim from the court’s own text.

How this was verified

Method: independent-origin adjudicated record, checked on 2026-08-24. The primary source is the full text of the OLG Hamm judgment of 12 May 2026 (Az. 4 UKl 3/25), published by the North Rhine-Westphalia justice portal (NRWE), from which the chatbot transcript, the attribution holding, the 250,000-euro per-violation fine ceiling, the 260.00-euro costs award and the admission of appeal to the BGH are all quoted verbatim. Two independent secondaries carry the same facts firsthand: Legal Tribune Online (dpa) and the case note of the Wettbewerbszentrale, the German competition-enforcement association (which notes it was not a party to the case). All three were archived to the Wayback Machine on 2026-08-24. The one open item is durability, not accuracy: the decision is not final and the Federal Court of Justice could still alter the attribution holding on appeal.

Related case files

• Air Canada was ordered to pay after its website chatbot invented a bereavement-fare policy — the same principle that a company owns what its bot tells a customer
• Cursor’s AI support bot ‘Sam’ invented a login policy that did not exist, triggering cancellations and a public apology — a chatbot confidently stating a false fact as company policy
• Pak’nSave’s ‘Savey Meal-bot’ generated dangerous recipes until warnings were added — another deployed consumer AI whose outputs became the operator’s problem

Sources

1. Oberlandesgericht Hamm (Justiz NRW / NRWE) · “Oberlandesgericht Hamm, 4 UKl 3/25” (full-text judgment of 12 May 2026) · 2026-05-12 · Tier 1 · https://nrwe.justiz.nrw.de/olgs/hamm/j2026/4_UKl_3_25_Urteil_20260512.html
2. Legal Tribune Online (dpa/LTO) · “OLG zur Haftung für irreführende KI-Chatbot-Aussagen” · 2026-05-12 · Tier 2 · https://www.lto.de/recht/nachrichten/n/olg-hamm-aerzte-haften-chatbot-falsche-facharztbezeichnung-ki-uwg
3. Wettbewerbszentrale · “OLG Hamm lässt Unternehmen für Aussagen seines Chatbots haften, Volltext verfügbar” · 2026-05-12 · Tier 2 · https://www.wettbewerbszentrale.de/olg-hamm-laesst-unternehmen-fuer-aussagen-seines-chatbots-haften-volltext-verfuegbar/
4. OECD.AI Incidents Monitor · “German Court Holds Doctors Liable for AI Chatbot’s False Medical Claims” · 2026-05-12 · Tier 2 · https://oecd.ai/en/incidents/2026-05-12-cfef