The RBI's Cautious Green Light for AI
The Reserve Bank of India is encouraging banks to accelerate their adoption of Artificial Intelligence, seeing it as a transformative force comparable to the liberalisation of the 1990s or the digital revolution of the 2010s. Speaking at the FIBAC 2026
conference, RBI Governor Sanjay Malhotra highlighted that AI is not just another project but a new way of running a bank. The central bank's stance isn't to contain AI as a risk, but to harness it responsibly as a powerful capability. However, this push for innovation comes with a strong emphasis on governance and oversight. The RBI has made it clear that while it supports progress, it will not tolerate a reckless approach, warning that carelessly deployed AI could create financial instability and new forms of social exclusion.
Unlocking AI's Potential in Banking
The RBI sees enormous potential for well-governed AI to reshape Indian finance for the better. A primary benefit is the expansion of financial inclusion. AI models can assess creditworthiness using alternative data like GST filings, utility payments, and digital footprints, bringing new-to-credit borrowers, gig workers, and small businesses into the formal banking system. For customers, AI promises better service through multilingual voice-based systems and personalised financial guidance. For banks, the benefits include enhanced productivity, automated back-office tasks, lower operational costs, and, crucially, more sophisticated fraud detection. The governor noted that as fraud moves at digital speeds, AI is essential to keep pace with criminals. AI-powered systems can also identify borrowers at risk of default earlier, allowing banks to offer support instead of initiating recovery actions.
The 'Black Box' Problem and Explainability
One of the biggest risks flagged by the RBI is the “black box” problem. Many advanced AI models arrive at decisions through processes that are not easily understood by humans. This lack of transparency is unacceptable when it affects customers' financial lives. Governor Malhotra stressed that if a loan is rejected by an AI, the bank must be able to explain why to the customer, auditors, and regulators. The phrase “the model decided” will not be considered an acceptable answer. This is the core of the demand for “Explainable AI” (XAI). Banks are being instructed to build the capacity to explain decisions that materially affect their customers. This means investing in technology and processes that make algorithmic decision-making transparent and auditable, ensuring fairness and accountability.
What 'Safe and Fair' AI Means for You
Beyond explainability, the RBI is focused on ensuring AI models are safe and fair. A major concern is algorithmic bias. If an AI is trained on historical data that contains biases related to geography, community, or occupation, the model can learn and even amplify those prejudices, leading to new forms of discrimination. The RBI insists that fairness must be a design requirement from the very beginning. Safety also encompasses robust data privacy and cybersecurity. With banks relying on vast amounts of data, protecting personal information is paramount, and the governor noted that simple compliance with the Data Protection Act is not enough. Furthermore, the AI systems themselves can become targets for cyberattacks. To counter this, the RBI has mandated that banks conduct rigorous stress-testing and “red-teaming” exercises to find and fix vulnerabilities before they can cause harm.
The Road Ahead: Accountability is Non-Negotiable
The RBI's new guidance places the ultimate responsibility for AI squarely on the banks themselves, not on third-party vendors or the algorithms. Every financial institution will be required to maintain a complete inventory of all AI systems it uses and establish a board-approved governance policy with clear lines of accountability. This marks a strategic shift, positioning model risk management as a core enterprise function rather than a niche technical concern. While some larger banks may develop AI in-house, many will rely on vendors. In these cases, the RBI is clear that accountability cannot be outsourced. Banks must ensure their vendors meet the same high standards of governance and transparency. This means a future where meaningful human oversight is a fundamental design principle, ensuring that a human can always intervene, explain, and override an AI decision when necessary.
