Skip to content

News

Regulation

Hamburg data watchdog: an LLM can be used lawfully if the system around it blocks data leaks

The Hamburg data protection authority published a case study on its LLMoin chatbot, built on an OpenAI model. It says the model is likely not anonymous, but input and output filters make its use acceptable.

The Hamburg Commissioner for Data Protection and Freedom of Information (HmbBfDI) published a case study on 5 November 2026 about LLMoin, the chatbot the Hamburg administration has used since December 2024. LLMoin runs on an OpenAI model.

The regulator does not claim that the model is free of personal data. It assumes that the model is, in all likelihood, not anonymous. It still considers the use acceptable because the system around the model prevents personal data from coming back out. The measures named in the case study are technical input and output filters, content filters, rules in the system prompt, usage rules for staff, and training. The authority also says the chatbot is an assistant only: staff still do the legal review, and the tool makes no decisions.

For customers of AI chat services and LLM APIs, the point is the reading of the GDPR. Whether a provider's model can hold personal data from training, and what the provider does to stop it leaking, is a question for the provider's own privacy policy and API data terms. The case study does not bind private providers, and it does not describe any crypto payment service.

Sources