P1023: Prüfkriterien für Reinforcement Learning (RL-Criteria)

TENDER DESCRIPTION

The project develops a catalogue of requirements for assessing the safety and trustworthiness of AI applications using reinforcement learning. The requirements are intended to cover as wide a range of applications as possible. In doing so, all techniques and forms of reinforcement learning are considered in principle, in particular also in connection with large and current generative models (keywords: RLHF, RLAIF). In the specific elaboration, the requirements are supplemented in each case by explanations, accompanying information, implementation instructions (e.g. explanations on the use of suitable test methods), and the like. The catalogue thus provides a basis for self-assessment and for testing relevant AI systems. The requirements created in the project are prepared in such a way that they are compatible with the BSI SdT library and fit into the existing BSI product portfolio for AI-specific test criteria. Specifically, the requirements are to be designed in the end as an additional module to BSI A5.

TENDER BRIEF

TENDER DOCUMENTS