'Watch out for biases while training AI'

MUMBAI: RBI deputy governor T Rabi Sankar on Tuesday urged safeguards in financial AI systems , stressing that oversight must always rest with humans. Speaking at the Global Fintech Fest here, he warned that the growing use of artificial intelligence in finance carries risks of bias, opacity, and weakened accountability.

"Bias in training data can entrench inequalities, and opaque models make explainability difficult, weakening regulatory scrutiny," he said.

He also highlighted ethical concerns, including misuse of behavioural data, privacy breaches, manipulative cross-selling, and social risks such as job displacement, while noting that excessive automation could heighten volatility and blur legal responsibility for errors.

Sankar launched three digital payment products at the event: UPI multi-signatory, small-value transactions using wearable glasses via UPI lite, and forex on Bharat Connect. UPI multi-signatory enables approval-based payments from joint accounts, improving efficiency and transparency.


The wearable integration allows small-value, voice-authenticated payments through smart glasses, easing seamless digital transactions. Forex on Bharat Connect links Clearing Corporation of India Limited 's FX retail with Bharat BillPay , allowing users to buy US dollars at real-time rates through bank apps.

Sankar elaborated on RBI's plans to strengthen protection against frauds and enhance oversight through technological solutions. Central to this is the forthcoming digital payments intelligence platform. "The platform will enable continuous monitoring, rapid anomaly detection, and timely intervention, ensuring resilience in our payments ecosystem," he said.

He also highlighted ongoing initiatives such as the Free AI Committee and the Mule Hunter tool, which aim to promote safety, fairness, and transparency in AI applications in finance.