Overview of Initiatives Suitable for Learning Assurance of AI-Based Military Products and Identified Challenges based on Use Cases Analysis

Machine learning (ML) can enhance military system performance and safety, crucial for future competitiveness. However, this shift from rule-based to data-driven approaches is not covered by current regulations and standards. To ensure reliability and safety, EASA and EUROCAE WG-114 introduced a learning assurance process for civil aeronautical ML-based systems. This paper explores challenges and specificities of applying the proposed learning assurance process for certifying and qualifying military ML-based systems using relevant use cases, focusing on the requirements identification and architecture definition.