AI, Earnings Management and Audit Committee

In the rapidly evolving landscape of 4inancial reporting, traditional corporate governance frameworks centered on monitoring human managerial decisions are increasingly shifting toward AI governance, which focuses on overseeing algorithmic design and automated data processing. This study investigates the empirical effect of arti4icial intelligence (AI) implementation on earnings management and examines the moderating role of audit committee oversight mechanisms. Utilizing a sample of 217 manufacturing 4irms listed on the Indonesia Stock Exchange over the 2023–2024 period (comprising 434 4irm-year observations), this research employs 4ixed-effects panel regression analysis with robust controls for 4irm-level heterogeneity. AI implementation is operationalized through a comprehensive textmining disclosure index in annual reports, while earnings management is evaluated across two distinct dimensions of accrual and real earnings management. The empirical 4indings reveal that AI implementation exerts a statistically signi4icant negative impact on both accrual and real earnings management, con4irming that AI adoption serves predominantly as an automated internal control and monitoring-enhancing mechanism that mitigates managerial opportunistic behavior rather than as an instrument for algorithmic manipulation. Crucially, moderation analysis demonstrates that both audit committee technology expertise and meeting frequency signi4icantly amplify this negative relationship, indicating that the ef4icacy of AI in curbing 4inancial statement distortions is strictly contingent upon governance quality. The primary originality of this research lies in extending agency theory into the digital era by illustrating that digital transformation mitigates information asymmetry only when backed by strong governance that bridges algorithmic opacity.