This paper examines the use of artificial intelligence in financial-market supervision, focusing on the benefits, risks, and legal safeguards associated with the integration of automated systems into the activities of supervisory authorities. While the use of artificial intelligence by financial institutions has already attracted significant regulatory attention, its deployment by public supervisory bodies remains comparatively underexplored. The study analyses the framework established by the EU AI Act, with particular regard to the potential classification of AI systems used in financial supervision as high-risk systems and to the resulting obligations concerning risk management, transparency, traceability, and human oversight. It argues that human oversight cannot be reduced to a mere formal requirement but must enable supervisory authorities to understand, contextualise, and, where necessary, challenge the probabilistic outputs generated by artificial intelligence systems. Drawing on a case study concerning market manipulation under Swiss law, the paper further shows that the duty to state reasons can operate as a central safeguard of the intelligibility, accountability, and legitimacy of decisions adopted with the support of artificial intelligence.