MMOOZZGG.com

Glossary

What is AI in banking

In short

AI in banking means models that assess credit risk, find fraudulent transactions, answer customers and check documents.

AI in industries · AI in banking

MMOOZZGG.com editorsChecked: 1 min read

For a child

Imagine a class piggy bank watched over by a student on duty. If someone suddenly tries to take all the coins at once, the student on duty gets suspicious and calls the teacher. A bank has the same careful helper: it notices strange purchases and stops them until the card owner confirms it was really them. That is why you must never tell anyone the codes from bank messages.

In plain words

A bank makes thousands of similar decisions every day: whether to give a loan, whether to let a payment through, what to answer a customer. Models learn from history: from past borrowers they estimate the chance of non-repayment, and from past transactions they learn to tell a normal purchase from a suspicious one. Chatbots and voice assistants answer questions about cards and transfers, and language models review contracts and statements. For example, if someone tries to pay for expensive electronics in another country with a card at night, the system pauses the transaction and asks the customer to confirm it.

For business

Finance has worked with models for a long time: scoring and fraud protection appeared long before chatbots. The new generation of tasks is tied to language models: support, document review, help for staff. The regulator sets the frame. On 9 July 2025, the Bank of Russia, the country’s central bank and financial regulator, issued a Code of Ethics for AI in the financial market. It recommends giving customers a way to refuse dealing with AI and to ask for a decision to be reviewed by an employee, telling them when AI is used, and labeling generated content. This leads to project requirements: explainable decisions, data kept inside the bank’s own perimeter and a log of checks. That is why banks often deploy language models on their own servers.

For a specialist

Classic tasks: credit scoring on gradient boosting and logistic regression, fraud detection on the transaction stream, anti-money laundering (AML) and churn forecasting; language models handle the new tasks in support and document analysis. The Bank of Russia’s code of ethics rests on five principles: human-centricity, fairness, transparency, safety, reliability and efficiency, and responsible risk management. It recommends not taking national, language or racial origin, membership in parties and public associations, or religion into account in decisions, and checking datasets for completeness and accuracy. In its report of 14 November 2024, the Financial Stability Board named among the vulnerabilities dependence on a few suppliers, synchronized behavior of market participants, cyber risks, model risks and data quality. In the EU, the AI Act classes creditworthiness assessment of individuals as high risk.

Example

A bank connected a language model that prepares a short summary of a borrower’s document package for an employee. The employee makes the loan decision, and the model runs on the bank’s servers and sends no data outside.

Related terms

Questions and answers

How do banks use AI? To assess credit risk, find fraudulent transactions, answer customers in chats and by phone, and check documents. Language models more and more often help staff prepare summaries and replies.
Can a customer refuse to deal with AI at a bank? The Bank of Russia’s code of ethics recommends giving customers a way to talk to an employee and to ask for a review of a decision made with AI. The document is a recommendation, and each organization sets its own procedure.

Sources

Not sure whether your company needs this?

AI solutions map
Contact us