The German Bundestag plans to make its own AI chatbot available to MPs and staff by the end of 2026. It is intended to support common knowledge and writing tasks, such as transcribing spoken content, editing and improving texts, preparing presentations and sorting routine enquiries. Another planned use is the categorisation of enquiries from members of the public, which have increased in number recently.
A key objective is to replace the existing use of external AI services with a more controlled alternative. The technical solution, developed by a Berlin-based start-up, can integrate different large language models. At the same time, personal data is to be masked automatically before requests are passed on to the respective AI model. The usage rules are based on the AI Act: entering personal data is prohibited, and AI-generated texts, including speeches, must be labelled accordingly.
The governance model is particularly relevant for companies and other organisations. Users receive individual accounts, must complete training before access is granted and are required to sign usage guidelines. Responsibility for the results remains with the users.
The project therefore illustrates one possible approach to the professional use of AI: rather than leaving employees to rely on freely available tools of their own choosing, organisations can provide centralised applications in which data protection, access rights and usage rules are built in from the outset.
By Maria Kleiner
Source: Heise, 23 September 2026, by Stefan Krempl