he purpose of the publication was to create a model that, based on the annual financial statements, identifies the risk of significant financial irregularities occurring in the enterprise. These irregularities may relate to different types of financial fraud that do not necessarily affect the annual financial statements. A characteristic feature of irregularities is that they are large-scale and will have a drastic impact on the company’s reputation. The results of the research show that machine learning and artificial intelligence algorithms were able to learn to recognize patterns of such scams and can detect them very effectively. An element of the novelty of the presented research is that it shows the possibility of training algorithms to recognize fraud based on information that is often not related directly to the observed fraudulent activities. The practical importance of research is the possibility of using the model in the decision-making process in the enterprise. The model allows assessing the risk that a potential business partner may commit financial fraud, which requires careful examination of the integrity of such an enterprise.