.png){kind=logo}
Reserve Bank of India Governor Sanjay Malhotra urged banks to use artificial intelligence to widen access to formal credit. AI could assess borrowers whom traditional systems struggle to evaluate through standard credit records.
The proposal does not ask software to reverse a loan officer’s rejection. Instead, alternative data could make previously overlooked applicants assessable. Still, Malhotra says banks must control the technology and answer for every decision.
Speaking at the FIBAC 2026 conference in Mumbai, Malhotra urged banks to shape their own AI strategies. He described AI as “a capability to be responsibly harnessed and not merely as a risk to be contained.” The RBI published his complete remarks.
Traditional underwriting often relies on credit scores, repayment histories, income records, and formal business accounts. Many potential borrowers lack enough information, even when they earn regular income or run viable businesses.
Malhotra said AI models can review alternative data, including cash flows, GST filings, utility payments, and digital platform records. Banks could use these records to assess repayment capacity and bring more customers into the formal lending system.
India’s digital systems can support that work. Aadhaar, UPI, DigiLocker, Account Aggregator, ONDC, and the Unified Lending Interface provide payment and data links. Malhotra said AI could change financial decisions as UPI changed digital payments.
Malhotra also warned banks about opaque models that cannot clearly explain their recommendations. A borrower should receive a clear reason when a bank denies credit. Auditors, bank boards, and the RBI also need enough information to review each decision.
Poor training data can repeat bias against certain places, occupations, or communities. Banks must also manage privacy, cyber threats and dependence on a small group of technology providers. A fault in a widely used model could spread across several lenders.
AI can support more than loan approvals. Predictive models can identify borrowers who may soon miss payments. Banks could then offer help before starting recovery. AI tools can also improve liquidity forecasts, stress tests, fraud detection and customer service.
Meanwhile, voice tools in Indian languages could help customers who face language or literacy barriers. AI-assisted staff could answer questions, identify suitable products and handle complaints more quickly.
The RBI expects banks to keep a complete inventory of every AI system they use, including tools inside vendor products. Boards should approve AI policies and assign clear responsibility for customer and business outcomes.
Banks should test AI systems before deployment and repeat those checks during use. They should challenge models with difficult cases, monitor their output, and keep staff ready to override recommendations.
Malhotra said banks cannot transfer blame to software providers or algorithms. “The model decided” cannot serve as an answer when customers, auditors, or the RBI question a decision. Human oversight must stay within lending and fraud controls.
Each bank will need an AI plan that matches its customers, risk limits, and ability to supervise technology. Smaller lenders may rely on outside providers, but they still must understand and control those systems.
The RBI’s approach follows recommendations from its FREE-AI Committee and draft guidance on model risk management. The central bank plans to use proportionate rules, consultation, evidence and regular reviews as AI systems change.
