The Reserve Bank of India has proposed draft rules requiring banks and other regulated entities to strengthen oversight of artificial intelligence and machine-learning models, including the ability to immediately override or deactivate systems that produce harmful or erroneous outputs. The draft guidance is aimed at building a formal model risk management framework across institutions using AI in financial decision-making.
Under the proposal, regulated entities would need board-approved policies covering all models, whether built in-house, sourced from vendors or created through a mix of both. Reuters reported that the framework would also require stronger controls, model inventories and risk-based review of AI systems. The guidance is open for public consultation until July 24, 2026.
The RBI has also pushed human oversight into the process by warning about automation bias, or the risk that staff may rely too heavily on machine output without independent judgment. For customer-facing AI systems, institutions would need to tell customers they are interacting with an AI system and provide an option to reach a human representative.
Explainability is another core requirement in the draft. Banks would need to define thresholds for how clearly a model’s output can be explained, especially in material decisions such as credit-related outcomes. The proposal also says third-party models are not exempt from validation, and institutions must independently test them before deployment.
The draft signals a more formal regulatory stance toward AI in Indian finance, with governance responsibilities pushed to the board level and controls expected to follow the risk level of each model.
Read Article: Tata Reports Extensive Damage at Delhi Data Centre as Clients Fear Decades of Data Loss
