The article, based on an analysis of the current legislation of Russia and the experience of legal regulation of developed countries, draws conclusions about the need to improve the concept of artificial intelligence technologies based on a criteria-based approach. The author proposed a classification of AI technologies according to the intended purpose of use: used for public purposes; used for commercial purposes; AI technologies used for non-commercial purposes. In banking, AI technologies are always used for commercial purposes, and in the financial sector, developers and people using AI are the same. The author concludes that a presumption of guilt has been established for financial organizations using AI technologies in relations with consumers. The proposed classification may be useful in connection with the creation of a system of regulation and supervision, since the current measures of responsibility of financial organizations should be considered insufficient.