Reducing software-related risks in autonomous ship’s development: insights from a cross-industry review

The rapid development of artificial intelligence (AI) has accelerated the advancement of autonomous systems by enabling higher levels of automated decision-making and control. In the maritime sector, autonomous ships are increasingly regarded as an important future direction for shipping. This study reviews current approaches to software safety and investigates how AI-related risks are addressed in autonomous-system development across different industries, with particular attention to autonomous ships. Following the PRISMA guidelines, 389 research articles from the Web of Science Core Collection and 76 AI-related incident reports from transportation industries were analysed. In addition, 11 maritime regulatory and classification society guidelines were evaluated using a structured qualitative coding scheme to examine software safety and AI-specific provisions. The review identifies several challenges associated with AI-driven systems, including adversarial attacks, tail-risk scenarios, limited causal understanding, and uncertainty in decision-making. Existing maritime frameworks remain insufficient in software reliability, verification, validation, and uncertainty-aware risk management for complex operational environments. Comparisons with the automotive and other transportation sectors highlight the need for adaptive stress testing, scenario-based simulation, formal verification, digital twins, and AI-specific risk assessment methods to improve the resilience and reliability of AI-driven autonomous ship systems.