The Ministry of Electronics and Information Technology (MeitY) has proposed mandatory human-in-the-loop interventions for agentic AI payments exceeding certain financial thresholds, according to the Digital Threat Report 2025-26 by CERT-In. This measure aims to ensure a human monitors or intervenes in AI-driven payment workflows, maintaining full audit trails for accountability.
CERT-In, the cyber wing of the IT ministry, outlined the proposal to mandate human oversight in agentic AI actions involving payments above defined limits. The report emphasizes that humans should provide feedback, correct errors, or make financial decisions to improve AI accuracy and reliability. This proposal coincides with the National Payments Corporation of India (NPCI) developing a Unified Agent Protocol to enable AI agents to make UPI payments, though details of the protocol remain undisclosed.
The move addresses concerns about trust in agentic commerce, as highlighted by MediaNama founder Nikhil Pahwa, who noted that without guardrails, recourse, and reversibility, agentic payments could lead to misuse. Pahwa cited examples from the US where AI agents made questionable purchases, underscoring the need for regulatory safeguards in India’s emerging AI payment systems. The proposal reflects growing attention to balancing innovation with security in digital payments.
The Digital Threat Report 2025-26 from CERT-In formally recommends these human-in-the-loop controls, marking a significant step toward regulating AI-driven financial transactions in India. NPCI’s ongoing development of the Unified Agent Protocol is expected to clarify operational details in the near term.