India's IT Ministry Proposes Mandatory Human Oversight in AI Payments

India's Ministry of Electronics and Information Technology (MeitY), through its cyber security unit CERT-In, has proposed mandating human oversight for artificial intelligence agents performing payment transactions above a defined financial limit. The suggestion, detailed in the Digital Threat Report 2025-26, aims to introduce safeguards as AI-driven financial activities evolve.

The inclusion of a "human-in-the-loop" mechanism means that a human will monitor or intervene in the AI's workflow, providing feedback or making decisions to enhance accuracy and prevent misuse. This proposal comes at a time when the National Payments Corporation of India (NPCI), which manages the Unified Payments Interface (UPI), is reportedly exploring ways to enable AI agents to make UPI payments.

Nikhil Pahwa, founder and editor of MediaNama, emphasized the need to build trust before enabling agentic UPI payments. He cited potential issues such as fraud and unauthorized transactions, drawing parallels to incidents in other markets where agents have made questionable purchases. Pahwa advocates for robust security features, including distinct agent identifiers and user-defined transaction limits.

Pahwa further suggested that AI agents should utilize digital wallets rather than bank accounts and be assigned unique, delegable handles. This approach could improve accountability and traceability. The ongoing development of agentic commerce protocols by companies like Coinbase and Pine Labs highlights the industry's direction, though questions surrounding data privacy and security remain.