Does AI contribute to systemic risk reduction in banks- Evidence from double-debiased machine learning

• This is the first study to examine the impact of artificial intelligence (AI) on bank systemic risk. • The study employs the Double-Debiased Machine Learning (DDML) model, which enables more accurate causal inference. • AI is found to reduce the level of bank systemic risk, with a more pronounced effect on tail systemic risk. • Mechanistic analysis reveals that AI facilitates the reduction of bank systemic risk by improving banks’ risk perception capabilities and promoting income diversification. • Banks are advised to proactively advance the integration of AI technologies into their risk management practices, so as to enhance their operational stability and better fulfill their role in serving the real economy. This study examines the impact of artificial intelligence (AI) and bank systemic risk (BSR) using double-debiased machine learning method. Based on data from 2010 to 2024, this paper finds that AI can reduce BSR, and has better performance in tail risk, this remains robust across several robustness checks. Mechanism analysis reveals that AI reduces BSR by enhancing risk perception capabilities and promoting income diversification. Moreover, heterogeneity analysis shows that this effect is stronger in state-owned banks and banks with a higher level of regional fintech level. Further analysis concludes that this effect can promote liquidity creation. This study proposes policy recommendations to maintaining banking stability and ensuring the operation of banking functions.