Testing Criteria for Reinforcement Learning (RL-Criteria)

Catalog of requirements for assessing the security and trustworthiness of AI applications using Reinforcement Learning (RL): Consideration of all RL techniques and manifestations, including large and current generative models (RLHF, RLAIF); requirements supplemented by explanations, information, and implementation guidelines (e.g., suitable test methods); the catalog serves as a basis for self-assessment and testing of AI systems; compatibility with the BSI's SdT library and integration into the BSI's existing product portfolio for AI-specific testing criteria; final design as a supplementary module to BSI A5.

Tender documents

Additional details