Fact sheet: ChatAI (Academic Cloud)

Description

Chat AI is an AI chatbot from Academic Cloud, operated by GWDG (Gesellschaft für wissenschaftliche Datenverarbeitung mbH Göttingen). The service provides access to various large language models (LLMs) via a web-based interface, including self-hosted open source models such as Meta LLaMA 3, Mistral and Qwen, as well as external models from OpenAI. Users can interact with the models via text message, audio recording or text file.
TU members can access the service via the Academic Cloud's federated login

Use in teaching

• Using AI as a sparring partner Chat AI is suitable as a dialogue-based sparring partner for teachers and students: they can develop and structure ideas, have draft texts revised, check arguments or receive feedback on technical issues. The adjustable system prompt allows the model to be tailored to specific roles or tasks – for example, as a critical proofreader, exam simulator or brainstorming partner. The selection of different models makes it possible to compare answers and experience the strengths and limitations of generative AI in a teaching context.

Stephanie Wichmann from the ZLL will be happy to advise you on the didactic use of the tool.

Technical requirements

Chat AI is a web application and only requires an up-to-date browser. No installation is necessary. An Academic Cloud account is required for access. TUHH members can log in via the federated login with their usual TUHH access data (Kerberos ).

Further links

• Chat AI – Direct access (login required)
• GWDG documentation on Chat AI
• Overview of available models
• Privacy policy (GWDG)
• Academic Cloud – Information page on Chat AI

Cost

The use of self-hosted open source models is currently free of charge for all TU university members. Commercial OpenAI models (ChatGPT) are excluded. These are currently only available to public universities and research institutions in Lower Saxony and the Max Planck Society (as of April 2026).

Data protection

GDPR-compliant AI-Tool

The self-hosted open source models run entirely on the GWDG's HPC infrastructure in Göttingen. User inputs and responses are not stored and are discarded immediately after processing. Chat histories only exist locally in the browser and are deleted when the session is closed. When using the external OpenAI models, requests are forwarded to Microsoft servers in European data centres – personal information in the requests is transmitted to the external service provider. For billing purposes, the GWDG itself only stores the date, user ID and length of the requests. Further details can be found in the GWDG's privacy policy and on the Academic Cloud information page.

A service provided by the Centre for Teaching and Learning (ZLL) at TU Hamburg.