The implementation of the Reserve Bank of India's (RBI) new FREE-AI framework will require significant investment, new governance structures, and sector-wide capacity building, experts said even as they added that it is a necessary step in safeguarding trust and accountability.
The framework, which calls for fairness, robustness, efficiency, and explainability in AI systems, is aimed at reshaping how banks and financial services deploy artificial intelligence.
From an implementation perspective, experts believe that cultural change is critical. Ajay Trehan, founder and CEO of AuthBridge, said, "Explainability means every AI-driven decision must be traceable to logic a human can understand, while fairness means ensuring no demographic is unduly disadvantaged." He added that the RBI's move makes responsible AI a compliance requirement, extending it from just best practice.
While the upfront costs of infrastructure, training, and monitoring may pressure some banks, experts agree that long-term benefits will outweigh these challenges. Improved decision quality, reduced compliance risk, and increased customer trust are expected to follow. "The hardest part is rarely the technology; it's cleaning legacy data and driving cultural change so that fairness and transparency are built in from day one," said Trehan.
According to Karthik Pasupathy, partner, Financial Services Risk Consulting, EY India, auditing an AI system under FREE-AI is not just about testing algorithms.
"Auditing involves evaluating governance processes, use case approvals, involvement of risk and compliance functions, and documenting decisions," he said. Broader evaluations involve running models through rare or extreme cases to ensure they adhere to laws and ethical norms. He noted that while techniques like SHAP and LIME can explain complex models, fairness eventually depends on the quality of training data. "Data cleaning is often the hardest part," added Pasupathy, pointing to the challenge of removing historical biases.
Most Indian banks are still in the early stages of AI adoption, transitioning from rule-based to simpler machine learning models.
The framework, which calls for fairness, robustness, efficiency, and explainability in AI systems, is aimed at reshaping how banks and financial services deploy artificial intelligence.
From an implementation perspective, experts believe that cultural change is critical. Ajay Trehan, founder and CEO of AuthBridge, said, "Explainability means every AI-driven decision must be traceable to logic a human can understand, while fairness means ensuring no demographic is unduly disadvantaged." He added that the RBI's move makes responsible AI a compliance requirement, extending it from just best practice.
While the upfront costs of infrastructure, training, and monitoring may pressure some banks, experts agree that long-term benefits will outweigh these challenges. Improved decision quality, reduced compliance risk, and increased customer trust are expected to follow. "The hardest part is rarely the technology; it's cleaning legacy data and driving cultural change so that fairness and transparency are built in from day one," said Trehan.
According to Karthik Pasupathy, partner, Financial Services Risk Consulting, EY India, auditing an AI system under FREE-AI is not just about testing algorithms.
"Auditing involves evaluating governance processes, use case approvals, involvement of risk and compliance functions, and documenting decisions," he said. Broader evaluations involve running models through rare or extreme cases to ensure they adhere to laws and ethical norms. He noted that while techniques like SHAP and LIME can explain complex models, fairness eventually depends on the quality of training data. "Data cleaning is often the hardest part," added Pasupathy, pointing to the challenge of removing historical biases.
Most Indian banks are still in the early stages of AI adoption, transitioning from rule-based to simpler machine learning models.
