Риски внедрения агентного искусственного интеллекта в финансовый мониторинг: подходы к классификации и оценке

Адилов Р.И.1 , Ступникова Е.А.2
1 Финансовый университет при Правительстве Российской Федерации, Москва, Россия
2 Российский университет транспорта (МИИТ), Москва, Россия

Статья в журнале

Экономическая безопасность (РИНЦ, ВАК)
опубликовать статью | оформить подписку

Том 9, Номер 11 (Ноябрь 2026)

Цитировать эту статью:

Аннотация:
Банк России относит агентные системы искусственного интеллекта к главным направлениям развития технологии, однако их применение в противодействии отмыванию преступных доходов сдерживается отсутствием методики оценки связанных с ними рисков. В отличие от моделей машинного обучения, которые выставляют оценку и оставляют решение сотруднику, агент самостоятельно собирает сведения, анализирует операции и выполняет действия с правовыми последствиями для банка и клиента. Цель исследования – разработать классификацию рисков внедрения агентного искусственного интеллекта в финансовый мониторинг кредитных организаций и метод их оценки, учитывающий полномочия агента. На основе сопоставления документов регуляторов, международных рамочных документов и научных публикаций выделено 14 рисков в пяти группах по источнику возникновения. Построена шкала из четырех уровней автономии агента, а матрица «вероятность – тяжесть последствий» дополнена коэффициентом автономии. Расчет показал, что переход от агента, готовящего заключения для сотрудника, к агенту, самостоятельно исполняющему решения, повышает среднюю оценку риска на 50,6% и переводит галлюцинации модели и косвенные внедрения вредоносных инструкций в категорию критических. Обосновано, что решения об отказе в проведении операций и о направлении сведений в Росфинмониторинг на текущем этапе должны оставаться за сотрудником банка. Результаты применимы при разработке внутренних правил использования искусственного интеллекта в системах финансового мониторинга.

Ключевые слова: агентный искусственный интеллект, финансовый мониторинг, противодействие отмыванию доходов, модельный риск, уровни автономии, кредитные организации

JEL-классификация: O17, O36, G14, D81, E44

JATS XML

Источники:

ИСТОЧНИКИ:

1. Акопов, В. С. Управление внедрением искусственного интеллекта в центральных банках с учётом аспектов развития и безопасности / В. С. Акопов // Экономическая безопасность. – 2025 – Т. 8, № 2. – С. 343-362. – DOI 10.18334/ecsec.8.2.122748 – EDN SIUWXD.

2. Бабанская, А. С. Развитие инструментов мониторинга криптовалютных транзакций для целей финансовой разведки / А. С. Бабанская, А. М. Иванова // Управление финансовыми рисками. – 2026 – Т. 22, № 1. – С. 95-110. – DOI 10.18334/ufr.22.1.124685 – EDN OTOGRK.

3. Болонина, С. Е. Искусственный интеллект в обеспечении безопасности финансового сектора: превентивные меры и управленческие решения / С. Е. Болонина, М. Ю. Лев // Экономика, предпринимательство и право. – 2025 – Т. 15, № 6. – С. 4349-4364. – DOI 10.18334/epp.15.6.123355 – EDN PMYJLH.

4. Болонина, С. Е. Технологические аспекты цифровизации российского банковского бизнеса / С. Е. Болонина, А. А. Орлова // Экономическая безопасность. – 2025 – Т. 8, № 5. – С. 1205-1228. – DOI 10.18334/ecsec.8.5.123359 – EDN PDGQEL.

5. Ермоловская, О. Ю. Предотвращение мошеннических транзакций в финансовой сфере / О. Ю. Ермоловская // Экономическая безопасность. – 2025 – Т. 8, № 4. – С. 923-942. – DOI 10.18334/ecsec.8.4.123240 – EDN KJHCAW.

6. Казанцева, С. Ю. Практика применения и перспективы развития технологий искусственного интеллекта и робототехники в сфере финансового контроля / С. Ю. Казанцева, Д. А. Казанцев // Вопросы инновационной экономики. – 2021 – Т. 11, № 2. – С. 553-564. – DOI 10.18334/vinec.11.2.112073 – EDN DLLGHI.

7. Кашеварова, Н. А. Анализ современной практики применения технологии искусственного интеллекта в финансовой сфере и его влияния на трансформацию финансовой экосистемы / Н. А. Кашеварова, Д. А. Панова // Креативная экономика. – 2020 – Т. 14, № 8. – С. 1565-1580. – DOI 10.18334/ce.14.8.110708 – EDN SFEXBC.

8. Концепция искусственного интеллекта в деятельности центральных банков: институциональные возможности / М. Ю. Лев, А. И. Болонин, И. Б. Туруев, Ю. Г. Лещенко // Экономическая безопасность. – 2024 – Т. 7, № 4. – С. 781-808. – DOI 10.18334/ecsec.7.4.120831 – EDN QYSQDS.

9. Лещенко, Ю. Г. Финансовый мониторинг как механизм обеспечения экономической безопасности Российской Федерации / Ю. Г. Лещенко // Экономическая безопасность. – 2019 – Т. 2, № 4. – С. 371-384. – DOI 10.18334/ecsec.2.4.110116 – EDN IWLANT.

10. Лукичёв, П. М. Агенты искусственного интеллекта и риски их применения / П. М. Лукичёв // Вопросы инновационной экономики. – 2025 – Т. 15, № 4. – С. 1345-1364. – DOI 10.18334/vinec.15.4.123758 – EDN RAAJBU.

11. Лукичёв, П. М. Экономика искусственного интеллекта и концепция «Принципал - Агент» / П. М. Лукичёв, О. П. Чекмарев // Вопросы инновационной экономики. – 2022 – Т. 12, № 2. – С. 1069-1082. – DOI 10.18334/vinec.12.2.114514 – EDN JKIFQQ.

12. Лукичёв, П. М. Риски применения искусственного интеллекта в краткосрочном периоде / П. М. Лукичёв, О. П. Чекмарев // Вопросы инновационной экономики. – 2023 – Т. 13, № 4. – С. 2443-2460. – DOI 10.18334/vinec.13.4.119359 – EDN SDOAGS.

13. Николаев, Д. А. Цифровые финансовые активы и противодействие отмыванию доходов: инфраструктурные решения и регуляторные вызовы / Д. А. Николаев // Экономическая безопасность. – 2025 – Т. 8, № 6. – С. 1619-1636. – DOI 10.18334/ecsec.8.6.123467 – EDN AZDGIT.

14. Тургаева, А. А. Современные подходы к оценке рисков отмывания денег и финансирования терроризма в условиях развития цифровых финансовых активов / А. А. Тургаева // Финансы: теория и практика. – 2025 – Т. 29, № 4. – С. 100-111. – DOI 10.26794/2587-5671-2025-29-4-100-111 – EDN LUMGBL.

15. Трехфакторная модель первичного финансового мониторинга клиентов коммерческих банков / А. В. Ханина, А. В. Зимовец, С. В. Петренко, Д. Ю. Михайлов // Управление финансовыми рисками. – 2026 – Т. 22, № 1. – С. 111-128. – DOI 10.18334/ufr.22.1.124528 – EDN WOFJGH.

16. Чувилкин, Н. А. Противодействие отмыванию доходов и финансированию терроризма как фактор обеспечения экономической безопасности экономики и общества / Н. А. Чувилкин // Экономическая безопасность. – 2021 – Т. 4, № 4. – С. 1237-1258. – DOI 10.18334/ecsec.4.4.113487 – EDN QDCTBK.

17. Amodei D., Olah C., Steinhardt J. [et al.] Concrete problems in AI safety // arXiv. - 2016. - arXiv:1606.06565. - URL: https://arxiv.org/abs/1606.06565

18. Canhoto, A. I. Leveraging machine learning in the global fight against money laundering and terrorism financing: An affordances perspective / A. I. Canhoto // Journal of Business Research. – 2021 – Vol. 131. – pp. 441-452. – DOI 10.1016/j.jbusres.2020.10.012 – EDN OYUFHW.

19. Chan A., Salganik R., Markelius A. [et al.] Harms from increasingly agentic algorithmic systems // Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency (FAccT '23). - New York: ACM, 2023. -  DOI  https://doi.org/10.1145/3593013.3594033

20. Greshake K., Abdelnabi S., Mishra S. [et al.] Not what you’ve signed up for: Compromising real-world LLM-integrated applications with indirect prompt injection//Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security (AISec ’23). - New York: ACM, 2023. - DOI https://doi.org/10.1145/3605764.3623985

21. Artificial intelligence for anti-money laundering: a review and extension / J. Han, Y. Huang, S. Liu, K. Towey // Digital finance. – 2020 – Vol. 2, Num. 3-4. – pp. 211-239. – DOI 10.1007/s42521-020-00023-1 – EDN XECKVE.

22. Куршэн Э., Балч Т., Берд Д. Агентный регулятор: риски ИИ в финансах и предложенная агентная структура управления // arxiv. — 2025. — arXiv:2512.11933. — URL: https://arxiv.org/abs/2512.11933. DOI https://doi.org/10.2139/ssrn.6970398

23. Deep Learning and Explainable Artificial Intelligence Techniques Applied for Detecting Money Laundering-A Critical Review / D. V. Kute, B. Pradhan, N. Shukla, A. Alamri // Ieee access. – 2021 – Vol. 9. – pp. 82300-82317. – DOI 10.1109/ACCESS.2021.3086230 – EDN UOVEQZ.

24. Parasuraman, R. Complacency and Bias in Human Use of Automation: An Attentional Integration / R. Parasuraman, D. H. Manzey // Human Factors: The Journal of the Human Factors and Ergonomics Society. – 2010 – Vol. 52, Num. 3. – pp. 381-410. – DOI 10.1177/0018720810376055

25. Parasuraman, R. A model for types and levels of human interaction with automation / R. Parasuraman, T. B. Sheridan, C. D. Wickens // IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans. – 2000 – Vol. 30, Num. 3. – pp. 286-297. – DOI 10.1109/3468.844354

26. A survey on large language model based autonomous agents / L. Wang, C. Ma, X. Feng [etc.] // Frontiers of Computer Science. – 2024 – Vol. 18, Num. 6. – pp. 186345. – DOI 10.1007/s11704-024-40231-1 – EDN XTFUTO.

27. Wooldridge, M. Intelligent agents: theory and practice / M. Wooldridge, N. R. Jennings // The Knowledge Engineering Review. – 1995 – Vol. 10, Num. 2. – pp. 115-152. – DOI 10.1017/s0269888900008122

Страница обновлена: 21.11.2026 в 19:05:08

 

 

Risks of adopting agentic artificial intelligence in financial monitoring: approaches to classification and assessment

Adilov R.I., Stupnikova E.A.

Journal paper

Economic security
Volume 9, Number 11 (November 2026)

Citation:

Abstract:
The Bank of Russia considers agentic artificial intelligence systems among the main directions of technological development, yet their use in anti-money laundering is held back by the lack of a method for assessing the related risks. Unlike machine learning models, which produce a score and leave the decision to an employee, an agent collects information, analyses transactions and takes actions with legal consequences for the bank and the customer. The aim of the study is to develop a classification of risks arising from the adoption of agentic artificial intelligence in the financial monitoring of credit institutions and an assessment method that accounts for the agent’s authority. Based on a comparison of regulatory documents, international frameworks and academic publications, 14 risks are identified and grouped into five categories by their source. A four-level scale of agent autonomy is built, and the likelihood–severity matrix is supplemented with an autonomy coefficient. The calculation shows that moving from an agent that prepares conclusions for an employee to an agent that executes decisions on its own raises the average risk score by 50.6% and makes model hallucinations and indirect injections of malicious instructions critical. It is substantiated that, at the current stage, decisions to refuse transactions and to report to Rosfinmonitoring should remain with a bank employee. The results are applicable to developing internal rules for the use of artificial intelligence in financial monitoring systems.

Keywords: agentic artificial intelligence, financial monitoring, anti-money laundering, model risk, levels of autonomy, credit institutions

JEL-classification: O17, O36, G14, D81, E44

References:

REFERENCES:

Akopov, V. S. (2025). Managing the implementation of artificial intelligence in central banks, including development and security issues. Economic security, 8(2), 343-362. https://doi.org/10.18334/ecsec.8.2.122748

Amodei D., Olah C., Steinhardt J. [et al.] Concrete problems in AI safety // arXiv. - 2016. - arXiv:1606.06565. - URL: https://arxiv.org/abs/1606.06565

Babanskaya, A. S., & Ivanova, A. M. (2026). Development of tools for monitoring cryptocurrency transactions for financial intelligence purposes. Financial risk management, 22(1), 95-110. https://doi.org/10.18334/ufr.22.1.124685

Bolonina, S. E., & Lev, M. Yu. (2025). Artificial intelligence in financial sector security: preventive measures and management decisions. Journal of Economics, Entrepreneurship and Law, 15(6), 4349-4364. https://doi.org/10.18334/epp.15.6.123355

Bolonina, S. E., & Orlova, A. A. (2025). Digitalization of the Russian banking business: technological aspects. Economic security, 8(5), 1205-1228. https://doi.org/10.18334/ecsec.8.5.123359

Canhoto, A. I. (2021). Leveraging machine learning in the global fight against money laundering and terrorism financing: An affordances perspective. Journal of Business Research, 131, 441-452. https://doi.org/10.1016/j.jbusres.2020.10.012

Chan A., Salganik R., Markelius A. [et al.] Harms from increasingly agentic algorithmic systems // Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency (FAccT '23). - New York: ACM, 2023. -  DOI  https://doi.org/10.1145/3593013.3594033

Chuvilkin, N. A. (2021). Anti-money laundering and combating the financing of terrorism as a factor of ensuring the economic security of the economy and society. Economic security, 4(4), 1237-1258. https://doi.org/10.18334/ecsec.4.4.113487

Ermolovskaya, O. Yu. (2025). Preventing fraudulent transactions in the financial sector. Economic security, 8(4), 923-942. https://doi.org/10.18334/ecsec.8.4.123240

Greshake K., Abdelnabi S., Mishra S. [et al.] Not what you’ve signed up for: Compromising real-world LLM-integrated applications with indirect prompt injection//Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security (AISec ’23). - New York: ACM, 2023. - DOI https://doi.org/10.1145/3605764.3623985

Han, J., Huang, Y., Liu, S., & Towey, K. (2020). Artificial intelligence for anti-money laundering: a review and extension. Digital finance, 2(3-4), 211-239. https://doi.org/10.1007/s42521-020-00023-1

Kashevarova, N. A., & Panova, D. A. (2020). Analysis of the current practice of applying artificial intelligence in the financial sector and its impact on the transformation of the financial ecosystem. Creative economy, 14(8), 1565-1580. https://doi.org/10.18334/ce.14.8.110708

Kazantseva, S. Y., & Kazantsev, D. A. (2021). The practice of applying and prospects for the development of artificial intelligence and robotics technologies in financial control. Russian Journal of Innovation Economics, 11(2), 553-564. https://doi.org/10.18334/vinec.11.2.112073

Khanina, A. V., Zimovets, A. V., Petrenko, S. V., & Mikhailov, D. Yu. (2026). Three-factor model of primary financial intelligence of commercial bank clients. Financial risk management, 22(1), 111-128. https://doi.org/10.18334/ufr.22.1.124528

Kute, D. V., Pradhan, B., Shukla, N., & Alamri, A. (2021). Deep Learning and Explainable Artificial Intelligence Techniques Applied for Detecting Money Laundering-A Critical Review. Ieee access, 9, 82300-82317. https://doi.org/10.1109/ACCESS.2021.3086230

Leshchenko, Yu. G. (2019). Financial monitoring as a mechanism of ensuring economic security of the Russian Federation. Economic security, 2(4), 371-384. https://doi.org/10.18334/ecsec.2.4.110116

Lev, M. Yu., Bolonin, A. I., Turuyev, I. B., & Leshchenko, Yu. G. (2024). The concept of artificial intelligence in the activities of central banks: institutional opportunities. Economic security, 7(4), 781-808. https://doi.org/10.18334/ecsec.7.4.120831

Lukichev, P. M. (2025). Artificial intelligence agents and the risks of their application. Russian Journal of Innovation Economics, 15(4), 1345-1364. https://doi.org/10.18334/vinec.15.4.123758

Lukichev, P. M., & Chekmarev, O. P. (2022). Economics of artificial intelligence and the principal-agent concept. Russian Journal of Innovation Economics, 12(2), 1069-1082. https://doi.org/10.18334/vinec.12.2.114514

Lukichev, P. M., & Chekmarev, O. P. (2023). Risks of the artificial intelligence application in the short term. Russian Journal of Innovation Economics, 13(4), 2443-2460. https://doi.org/10.18334/vinec.13.4.119359

Nikolaev, D. A. (2025). Digital Financial Assets and Anti-Money Laundering: Infrastructure Solutions and Regulatory Challenges. Economic security, 8(6), 1619-1636. https://doi.org/10.18334/ecsec.8.6.123467

Parasuraman, R., Sheridan, T. B., & Wickens, C. D. (2000). A model for types and levels of human interaction with automation. IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans, 30(3), 286-297. https://doi.org/10.1109/3468.844354

Parasuraman, R., & Manzey, D. H. (2010). Complacency and Bias in Human Use of Automation: An Attentional Integration. Human Factors: The Journal of the Human Factors and Ergonomics Society, 52(3), 381-410. https://doi.org/10.1177/0018720810376055

Turgaeva, A. A. (2025). Modern Approaches to Assessing ML/FT Risks in Countries with a High Level of Digital Financial Assets. Finance: Theory and Practice, 29(4), 100-111. https://doi.org/10.26794/2587-5671-2025-29-4-100-111

Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., Zhao, W. X., Wei, Z., & Wen, J. (2024). A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6), 186345. https://doi.org/10.1007/s11704-024-40231-1

Wooldridge, M., & Jennings, N. R. (1995). Intelligent agents: theory and practice. The Knowledge Engineering Review, 10(2), 115-152. https://doi.org/10.1017/s0269888900008122