Secure AI Services for Research and How to Make Your Data Depositable in a FAIR Way

This presentation introduces secure AI services for research purposes, focusing on the responsible use of generative AI to prevent sensitive or unpublished data from being transmitted uncontrollably to commercial platforms. It shows how the University of Mannheim's maKI service can be used to make research data depositable in a FAIR way. The first part presents survey results on AI use in research and studies: self-assessed knowledge of generative AI, frequency of use, areas of application, and the models used. The second part defines Secure AI Services (data protection, information security, scientific use), explains when to use them, and lists use cases such as brainstorming, summarizing, transcription, metadata creation and agentic coding. The third part presents concrete services, including the "ChatAI" LLM web service by GWDG (Gesellschaft für wissenschaftliche Datenverarbeitung mbH Göttingen) and the OpenAI-compatible API service "maKI" of the University of Mannheim. It also introduces the Traffic Light Protocol (TLP) for classifying data sensitivity. Topics: AI use in research Secure AI Services: definition, use cases, when to use them Multimodal LLMs in practice: ChatAI (GWDG), maKI (University of Mannheim) Making data depositable in a FAIR way using maKI Code: A Jupyter notebook demonstrating the use of maKI can be found on GitHub: https://github.com/UB-Mannheim/research-skills/tree/main/2026_Secure_AI_Services_for_Research