Credit organizations are the first line in the fight against illegal financial activities. They are designed to collect and analyze information about their clients and their financial transactions, and report any suspicious activity to the Federal Financial Monitoring Service (Rosfinmonitoring) in the event of detected suspicious activity . However, credit institutions can also be involved in illegal activities: knowingly covering up shadow schemes, or due to a weak internal control system. To successfully solve the problem of money laundering with the involvement of credit institutions, a systematic approach and scientific understanding of the empirically obtained results are required. Automation of the process of identifying unscrupulous credit institutions based on machine learning methods will allow regulatory authorities to quickly identify and suppress illegal activities. Damage as a result of crimes related to the withdrawal of bank assets can be incurred not only by the bank’s depositors and customers, but also by the state and bona fide participants in the banking business. The purpose of the study is to improve the efficiency of detecting unscrupulous credit institutions by regulatory authorities. A necessary tool for this can be typological analysis to identify the content side of the methods and trends of money laundering, as well as modern methods of data analysis and machine learning – to automate the identification of deviant banks. The application of typological analysis in economics and other sciences is considered. Various typologies of money laundering with the involvement of credit institutions are considered and systematized. A comparative analysis of the results of processing data on the activities of credit institutions by anomaly search methods is carried out – a one-class support vector machine algorithm and an anomaly detection algorithm based on the principal component method. It is concluded that the algorithm for searching for anomalies based on the principal component method has shown more accurate results compared to the one-class support vector machine algorithm. The above research results can be used by the Bank of Russia and Rosfinmonitoring to automate the identification of unscrupulous credit institutions. The results of the study can also be used by internal control services in the credit institutions themselves in order to self-check and prevent the bank from being involved in dubious schemes, which will increase the responsibility of the subjects of financial monitoring. The author sees the directions for further research in the approbation of the methodology in relation to other subjects of financial activity: professional participants in the securities market, microfinance organizations, insurance organizations.