The governor of India’s Reserve Bank wants the nation’s financial institutions to use AI to approve loans that human assessors would likely reject.
In a speech delivered yesterday at the FIBAC conference in Mumbai, Sanjay Malhotra said the regulator “sees AI as a capability to be responsibly harnessed and not merely as a risk to be contained,” then offered several reasons to support that statement.
The first reason is that banks currently find it hard to justify loans to first-time borrowers, gig workers, or small businesses that don’t keep formal books.
“AI models, trained on alternative data – cash flows, GST filings, utility payments, digital footprints – can extend the frontier of ‘bankable’ India considerably further than manual underwriting ever could, at a fraction of the marginal cost per loan,” Malhotra argued. That’s a reference to the fact that India, like many other developing nations, has many people who either lack access to banks entirely or are “underbanked,” meaning they use alternative and/or unregulated lenders that could come with high costs.
“Predictive models can identify borrowers on the cusp of default early enough to counsel rather than merely recover,” he added. “Used well, AI may be the most powerful accelerator to financial inclusion.”
One way the tech can make that happen is if Indian banks use AI to develop voice interfaces in local languages. India recognizes 14 major languages that are spoken by ten million or more residents, plus another eight languages felt to be an important part of the nation’s heritage. Literacy rates in rural areas remain below 80 percent. Malhotra therefore thinks AI can help more people to work with banks.
MORE CONTEXT
The governor added his view that “AI-enhanced credit risk models, liquidity forecasting, and scenario analysis allow banks – and, indeed, us, as the regulator – to see emerging stress earlier than lagging financial statements permit.”
Malhotra also thinks AI could improve Indian banks’ customer service. “A relationship manager assisted by an AI system that presents the right product, the right risk flag, can serve a higher number of customers more efficiently,” he said. “AI-assisted grievance redressal, and personalised financial guidance can enhance service quality to customers.”
The governor is not blind to perils posed by AI and devoted a section of his speech to what he described as “Risks we must keep firmly in view.”
The first of those risks is what he called “The black box problem.”
“Many advanced AI models – particularly deep learning and generative systems – do not readily explain their own reasoning,” he observed. “When an AI system recommends against extending credit to a small business, both the borrower and the regulator are entitled to know why.” Beyond courtesy to customers, Malhotra worries that opaque AI decision-making “makes it exceedingly difficult for auditors, boards, and the Reserve Bank to be confident that a model is doing what it was designed to do.”
He also worries that AI could perpetuate “biases against certain geographies, certain occupations, certain communities.”
Dependence on a cluster of tech companies also concerns the banking boss, who fears several AI providers could offer flawed systems that spread a risk across India’s banking system.
He therefore set down a marker.
“No matter how sophisticated the model, the responsibility for a bank's decisions rests with the bank, not with its algorithm,” he said. “‘The model decided’ can never be an acceptable answer to a customer, an auditor, or the Reserve Bank. Meaningful human oversight – the ability to explain, to intervene, and, where necessary, to override – must remain a design principle, not an afterthought.”
He also set the following expectations for Indian banks’ use of AI:
-
Maintain a complete inventory of every AI system in use – including those embedded in vendor products – so that neither you nor we are ever surprised by what is running inside your institution.
-
Establish board-approved AI governance policies, with clear accountability for outcomes, not merely for technology procurement.
-
Build the capacity to explain AI-driven decisions that materially affect a customer, particularly in lending and fraud outcomes.
-
Red-team and stress-test AI systems before deployment and periodically thereafter, just as you would stress-test any other material risk.
-
Preserve meaningful human oversight at every point where an AI system's error could cause material harm to a customer or to financial stability.
“The banks that will win in the AI era will not necessarily be the ones that adopt the most AI, or the fastest,” he concluded. “They will be the ones that adopt it with the deepest understanding of what they are deploying, the clearest accountability for its outcomes, and the strongest commitment to the customer's trust that has always been, and will remain, the true capital of Indian banking.” ®
