Artificial intelligence in safety-critical systems: a systematic review

Purpose This study is a systematic literature review of the application artificial intelligence (AI) in safety-critical systems. The authors aim to present current status according different AI techniques and propose some research directions insights promote its wider application. Design/methodology/approach A total 92 articles were selected for this through along with thematic analysis. Findings divided into three themes: interpretable method, explain model behavior reinforcement safe learning. Among techniques, most widely used are Bayesian networks (BNs) deep neural networks. In addition, given huge potential field, four future also proposed. Practical implications vital interest industry practitioners regulators domain, as it provided clear picture pointed out that have great potential. For those inherently appropriate use systems, can conduct in-depth studies validate encourage their industry. Originality/value first systems literature. It marks step toward advancing domain. paper has values term “safety-critical” improve phenomenon fragmentation.