At the online conference "Exams and Deception - From Cheat Sheets to the Use of Artificial Intelligence" on 7 October 2025 organised by the Association for the Promotion of German & International Science Law e.V., essential questions regarding generative AI in written exams, in particular seminar papers, were discussed with the administrative judges of the VG Berlin:
- Generative AI can be an unauthorised or permitted aid.
- Aids do not necessarily have to be regulated in the exams regulations.
- The scope of the permitted use of generating AI should ideally be communicated in the declaration of independence.
- Due to their susceptibility to errors and the lack of reference text, results from AI detectors alone cannot be a deciding factor in a cheating decision.
- A random interview on the use of AI, independent of suspicion, should be permitted without a separate provision in the exams regulations, provided that this does not constitute a second examination. The legal basis for this is Section 26 VwVfG (duty to cooperate).
The conference report will be published in December 2025 in the OdW.
AI-robust exam formats and AI-integrating exam formats are currently being evaluated (see AG Exam formats). If these represent a new type of exam not provided for in the exams regulations, the exam regulations must be amended. The adaptation of teaching and exams to new technologies, such as AI, can eliminate the susceptibility to cheating through generating AI and also solves the question of how exams can test media and methodological skills in addition to didactic questions that take into account the realities of work in the future.