India’s central bank wants AI to approve loans that humans would reject

India’s central bank governor, Sanjay Malhotra, has urged financial institutions to harness artificial intelligence (AI) for approving loans that human assessors might reject. Speaking at the FIBAC conference in Mumbai, Malhotra emphasized that AI should be responsibly harnessed rather than merely viewed as a risk. He presented several compelling reasons for this approach.

Firstly, banks struggle to justify loans for first-time borrowers, gig workers, and small businesses that lack formal financial records. AI models, trained on alternative data such as cash flows, GST filings, utility payments, and digital footprints, can extend banking opportunities to a larger population at a lower marginal cost per loan.

This is particularly relevant for India, where many individuals lack bank access or are underbanked, relying on high-cost alternative lenders.

AI's predictive capabilities can identify borrowers at risk of default early, allowing for timely counseling instead of just recovery efforts. Malhotra also highlighted that AI can facilitate the development of voice interfaces in local languages, given India's 14 major languages spoken by over ten million residents and another eight languages of cultural significance.

With rural literacy rates below 80 percent, AI can help more people engage with banks. Furthermore, AI-enhanced credit risk models, liquidity forecasting, and scenario analysis enable banks to detect emerging financial stress early, surpassing what traditional financial statements can reveal. This technology can also enhance customer service by assisting relationship managers in efficiently serving more customers and providing personalized financial guidance and grievance redressal.

However, Malhotra acknowledged the inherent risks of AI, particularly the "black box problem," where advanced AI models like deep learning and generative systems lack transparency in their reasoning. This opacity makes it challenging for auditors, boards, and the Reserve Bank to trust AI's decision-making processes. Additionally, he expressed concerns about AI perpetuating biases against specific regions, occupations, and communities.

Concerning AI dependence, Malhotra warned about the potential for flawed systems spread across India's banking sector, urging banks to retain ultimate responsibility for their decisions, ensuring meaningful human oversight remains paramount. He outlined expectations for Indian banks' AI use, including maintaining a comprehensive inventory of all AI systems, establishing board-approved AI governance policies, building capacity for explaining AI-driven decisions, conducting regular red-team and stress-testing, and preserving meaningful human oversight throughout the AI decision-making process.

Malhotra concluded that the banks thriving in the AI era will be those that adopt it with a deep understanding, clear accountability, and a strong commitment to customer trust.