The Algorithm Will See You Now
Indian banking is undergoing a quiet but radical transformation. The friendly loan officer behind a desk is increasingly being augmented, and in some cases replaced, by complex algorithms. Driven by the promise of efficiency, speed, and reaching millions
of new customers, banks are rapidly adopting Artificial Intelligence (AI) and Machine Learning (ML) to make credit decisions. This shift to automated credit scoring allows lenders to process vast numbers of applications in minutes, a feat impossible for human teams. For a credit-hungry nation, this appears to be a win-win: banks cut costs, and more people get faster access to loans. However, this high-speed revolution comes with a significant blind spot. While the front-end experience is seamless, the back-end logic is often a 'black box', leaving both customers and sometimes even the banks themselves in the dark about how a decision was reached.
When the Computer Says No
The core of the problem lies in accountability. When a person is denied a loan, they have a right to an explanation. With an algorithm, that explanation is often missing. The Reserve Bank of India (RBI) Governor, Sanjay Malhotra, recently issued a stern warning to banks, stating that blaming an algorithm for a decision is not an acceptable answer. He stressed that the ultimate responsibility must always lie with the bank, not the technology vendor or the model itself. This is crucial because automated systems, despite their veneer of objectivity, can be deeply flawed. They learn from historical data, and if that data reflects existing societal biases, the AI can perpetuate and even amplify them. This can lead to a form of 'digital redlining', where entire demographics or geographies are unfairly penalised based on proxy data they have no control over.
Regulation is Catching Up, But Gaps Remain
Regulators are not sitting idle. The RBI has introduced comprehensive guidelines for digital lending, focusing on transparency, data privacy, and grievance redressal. These rules mandate that lenders provide a clear Key Fact Statement (KFS) with all costs, and they strictly limit the data that lending apps can access on a borrower's phone. Furthermore, the RBI is pushing banks to establish board-approved AI governance policies and maintain a full inventory of all AI models in use. However, a recent study highlighted that existing frameworks, while strong on data privacy, may lack the specificity to enforce true algorithmic accountability and explainability. The 'black box' problem persists, where tracing the exact reason for an adverse decision remains difficult, making a consumer's right to correction almost theoretical.
The Path to Fair and Transparent AI
True accountability requires more than just rules; it demands a shift in design and oversight. The solution isn't to abandon AI, which the RBI itself encourages banks to adopt responsibly, but to build guardrails around it. A critical step is mandating 'explainable AI' (XAI), where models are designed to provide clear, human-understandable reasons for their outputs. Banks must also implement robust 'human-in-the-loop' systems, ensuring that any contested or high-stakes decision can be reviewed by a person with the authority to override the algorithm. Independent and regular audits for bias in these models should become standard practice, not an optional extra. The regulator's message is clear: outsourcing the technology does not mean outsourcing the responsibility. Banks remain 100% accountable for the outcomes.
