What is AI in finance?

AI can be used in finance for any task that turns large amounts of data into a decision or a text. The European Banking Authority (EBA) lists the use cases of AI in finance that it observes in EU banks.

These are examples of AI in banking. For investment firms, the European Securities and Markets Authority (ESMA) names customer support, fraud detection, risk management, compliance, and support for investment advice and portfolio management (ESMA, May 30, 2024).

Machine learning in finance is the branch of AI in which a model derives its rules from past data, for example from the repayment records of earlier borrowers. The developer chooses the data and the goal; the model sets the weights.

Banks are using machine learning in credit risk models. The EBA consulted on its use in internal ratings-based (IRB) models, with which a bank calculates its own capital requirement for credit risk, and published a follow-up report on August 4, 2023. The report also asked for clarifications on how the General Data Protection Regulation and the AI Act apply to such models.

Generative AI in finance is AI that writes text or code, such as the large language models behind chatbots. The Financial Stability Board (FSB) notes that with generative AI and large language models "the range of use cases has become more diverse" (FSB, November 14, 2024).

In the EBA survey, 55% of banks use general-purpose AI or agentic AI in consumer-facing processes. The most common uses are fraud alerts, help for call center staff and self-service answers for customers. In Switzerland, 91% of the institutions that use AI also use generative AI, and FINMA sees a growing dependence on large technology providers.

The benefits of AI in finance are faster analysis and lower costs; the main risks that supervisors name are model errors and dependence on a few providers. The Bank for International Settlements (BIS) summed up both sides in June 2024: "Benefits include improvements for lending and payments; risks include more sophisticated cyber attacks" (BIS, June 25, 2024).

The FSB lists four vulnerabilities: "third-party dependencies and service provider concentration; (ii) market correlations; (iii) cyber risks; and (iv) model risk, data quality and governance." The European Central Bank (ECB) adds a warning about shared models: "If a majority of financial institutions use the same or very similar foundation models provided by a few suppliers, it is likely that decisions based on AI will suffer from similar biases" (ECB, Financial Stability Review, May 2024).

AI in finance is regulated by the EU AI Act and by the financial laws that already apply to the firm. The AI Act lists AI systems that evaluate the creditworthiness of natural persons or set their credit score as high-risk, with the exception of systems that detect financial fraud (Annex III, point 5(b)). Regulation (EU) 2026/1744 of July 8, 2026 moved the start of these high-risk rules to December 2, 2027.

For investment services, ESMA expects firms that use AI to comply with MiFID II, including "their regulatory obligation to act in the best interest of the client". Explainable AI in finance is a model whose result a bank can explain to the client and to the supervisor, for example the reason for a refused loan. FINMA counts a lack of explainability among the model risks of AI.

In Germany and Austria the AI Act applies directly. For high-risk AI systems used by financial institutions, Article 74(6) makes the financial supervisor the market surveillance authority, unless the member state names another authority under Article 74(7). The financial supervisors are BaFin in Germany and the FMA in Austria.

BaFin published its guidance on ICT risks in the use of AI at financial entities on December 18, 2025. The guidance is not binding. It applies the rules of DORA to AI and follows the life cycle of a model from data acquisition to retirement.

For AI in finance, Frankfurt is where the ECB supervises the significant banks of the euro area. ECB Banking Supervision, at Sonnemannstrasse 20 in Frankfurt am Main, reported a strong increase in AI use cases at these banks between 2023 and 2024. In 2025 it held workshops with 13 banks using AI for credit scoring and fraud detection: decision tree-based models were mainly used for both, neural networks mainly for fraud detection (ECB, November 20, 2025).

Switzerland is outside the EU, so the AI Act does not apply there. FINMA published Guidance 08/2024 on governance and risk management when using AI on December 18, 2024. It names model risks such as a lack of robustness, correctness, explainability or bias, and observes that most institutions are "still in the early stages of development". This page gives no legal advice.