The Reserve Bank of India (RBI) insists that lenders must provide an explanation for why algorithms influencing lending decisions rejected a borrower's application. This emphasizes accountability and explainability in RBI's expectations regarding technologies used in lending.
Shirish Chandra Murmu, Deputy Governor of the RBI, speaking at the Global Fintech Festival in Mumbai on Friday, stressed that digital lending should not only make the loan issuance process faster. He noted that technology and improved data should help lenders find borrowers who previously lacked access to the formal financial system, making credit higher quality, not just faster.
Murmu stated that new, more complex questions arise for people, such as vendors with printed QR codes, who have entered the formal financial system. He posed questions: will they be offered loans under clear terms? Will anyone be able to explain why an algorithm rejected ten of their applications? How long will they be unable to trade if their account is frozen due to fraud detection systems? These questions, according to Murmu, are the essence of trustworthy innovation, and the answer to them will be more important than any implemented technology.
Furthermore, the Deputy Governor cautioned institutions against over-reliance on cryptographic systems. He explained that while quantum computing does not pose an immediate threat, payment infrastructure has long development cycles, and transitioning to new standards takes years. Murmu also mentioned the risk of 'collect now, decrypt later,' where encrypted information collected today could become accessible as computational power advances.
This issue is particularly relevant for millions of small traders in India who have transitioned to the digital payment system. The RBI's main concern is that algorithms may make statistically effective decisions that are not easily understandable. Murmu noted that well-designed models can identify creditworthy borrowers who would otherwise remain outside formal financing, but warned that the data used by these models may contain historical biases. He added that past relationships might not be preserved, the model might be economically unviable despite its statistical sophistication, and complex models can be difficult to explain to a client whose application was rejected.
The RBI's stance is also that automation does not absolve the lender of responsibility and transfers it to the technology. When the question of responsibility for a decision made or significantly influenced by an algorithm arises, Murmu replied: 'The answer cannot be the algorithm.' He emphasized that the responsibility lies with the regulated institution, and boards of directors and senior management must understand the models they deploy, their limitations, and the consequences of their use.
This principle applies to the entire digital lending ecosystem. According to the RBI's Digital Lending Guidelines of 2025, technology may change the delivery channel of credit, but it does not reduce the responsibility of the regulated lender. Borrowers must know who is providing them the loan, the cost, and the terms, and technology should expand customer choice, not steer borrowers toward unsuitable products. Even when operating through partnerships, the regulated entity remains responsible for the services provided on its behalf.