What RBI's FIBAC speech asks of Indian banks' AI programs
RBI Governor Sanjay Malhotra told banks on August 11 that responsibility for an AI decision stays with the bank. He named four priorities: an inventory of AI systems, a board-approved policy, explainable customer decisions and red-teaming. Here is what each one means in the systems.
Banking & payments2 min read
At FIBAC 2026 in Mumbai on August 11, RBI Governor Sanjay Malhotra told banks that AI is not a single technology to buy or a project to finish, and that responsibility for a decision stays with the bank. A bank cannot point to the algorithm or the vendor. He asked banks to treat four things as immediate priorities, and each one lands in a specific place in a bank's systems.
An inventory that includes the AI inside vendor products
The first priority is a complete inventory of AI systems, including AI embedded in vendor products. The models a bank built are the easy part. Harder to find are the scoring inside a loan origination package, the chatbot a contact-center vendor switched on in a release, the anomaly detection inside a payments switch.
In practice that means asking every vendor, in writing, which of their features use a trained model, what data it reads, and how a decision can be traced. It also means recording, for each model, which decisions it affects and who in the bank owns them.
A policy the board approves and someone owns
The second priority is a board-approved AI governance policy with clear accountability. A policy only matters if each system in the inventory has a named owner who answers for its results and can stop it.
Decisions a customer can have explained
The third is the ability to explain decisions made with AI that affect customers, especially in lending and fraud. That is a design requirement, not a document. The system has to store the inputs, the model version, the score and the reason codes at the moment of the decision, because a model retrained next quarter cannot reconstruct why it declined someone last month.
Explaining a decision later depends on what the integration with the banking system records, so the logging belongs in the first design, not in a later compliance phase.
Red-teaming before launch and after
The fourth is red-teaming and stress-testing before deployment and periodically after. For a credit or fraud model that means testing it on shifted data, on deliberately manipulated inputs and on the segments where historical data was thin. For a language model it means trying to make it reveal data, ignore instructions or act outside its permissions.
The risks behind the priorities
The Governor also listed the risks behind these priorities. They include decisions nobody can explain, bias carried over from historical data, dependence on a few models or vendors, adversarial attacks and the erosion of human judgment. He described human oversight as the ability to explain, intervene and override, and said it must be a design principle.
Questions
- What did the RBI Governor say about AI at FIBAC 2026?
- On August 11, 2026, Sanjay Malhotra said responsibility for AI decisions stays with the bank. He named four immediate priorities: an inventory of AI systems including those inside vendor products, a board-approved governance policy, explainable customer decisions in lending and fraud, and red-teaming before and after deployment.
- Does a bank need to inventory AI inside vendor software?
- Yes. The Governor named AI embedded in vendor products as part of the inventory. Ask each vendor which features use trained models, what data they read and how decisions are traced.
- What does human oversight of AI mean for a bank?
- As the Governor described it, the ability to explain a decision, intervene in it and override it, built into the system from the start.
Sources
- RBI Governor Sanjay Malhotra Sets Out AI Playbook for Indian Banks, The Kanal, August 13, 2026
More on banking & payments
All writingNovember 2026
RBI's API advisory: the controls banks need by December 3
RBI has reportedly told banks to tighten API-based payments as AI-assisted attacks were reported: a full API client inventory, audits of API clients, payout safeguards and transaction limits, with key deadlines on December 3, 2026. What each control means in the systems.
October 2026
Fraud scoring inside a core banking system, step by step
Fraud scoring pays off most when the score reaches the transaction before it is approved. Here is how to get it there. When we put real-time scoring and alerts inside a major financial institution's banking systems, fraudulent transactions fell by 25%.
September 2026
What does human oversight of AI mean in practice for a bank?
RBI Governor Sanjay Malhotra described human oversight as the ability to explain, intervene and override an AI decision. In a bank's systems that means three things: a record for every decision, a place where a person can step in, and a switch that stops the model without stopping the business.
Tell us which system the AI has to run inside
If that system is not ready for AI yet, tell us that too. An engineer will reply and say plainly whether we are the right firm for it.