MeitY proposes mandatory human-in-the-loop interventions in agentic AI payments
India's Ministry of Electronics and Information Technology (MeitY) proposes mandatory human oversight for agentic AI payments exceeding certain financial thresholds, as outlined in the Digital Threat Report 2025-26.
Intelligence analysis by Gemini 2.5 Flash
Amid reports that AI agents may soon facilitate UPI payments, MeitY's cyber wing, CERT-In, is advocating for 'human-in-the-loop' controls to ensure security and accountability. This move aims to build trust and provide recourse in the evolving landscape of AI-driven financial transactions, addressing concerns about potential fraud and misuse.
Imagine you have a super smart robot helper that can buy things for you, like ordering your favorite snack. India's government wants to make sure that for bigger purchases, a grown-up always checks with the robot before it spends your money. This is like a parent making sure you don't accidentally buy too many toys, keeping your money safe and making sure the robot doesn't make a mistake.
Analysis
MeitY's Regulatory Stance on AI Payments
India's Ministry of Electronics and Information Technology (MeitY), through its cyber wing CERT-In, has put forth a significant proposal to mandate 'human-in-the-loop' interventions for agentic AI payments. This recommendation, detailed in the Digital Threat Report 2025-26, specifically targets transactions exceeding defined financial thresholds. The core idea is to ensure that a human monitors or actively intervenes in AI workflows, providing feedback, correcting errors, or making critical financial decisions to enhance the accuracy and reliability of AI systems. This proactive regulatory approach underscores the government's intent to establish guardrails before widespread adoption of autonomous AI in sensitive financial operations.
The Rise of Agentic Payments in India
The proposal from MeitY comes at a time when the concept of AI agents making payments is gaining traction within India's digital economy. Reports indicate that the National Payments Corporation of India (NPCI) is actively developing a Unified Agent Protocol to enable AI agents to conduct UPI payments. This initiative, if successful, could revolutionize how transactions are executed, allowing for more autonomous and efficient commerce. Beyond NPCI, companies like Pine Labs have already launched agentic payment protocols, such as P3P built on UPI, and open-sourced their agentic OAuth framework. Even consumer-facing platforms like Zepto Cafe are testing AI agents capable of placing orders for users, signaling a broader industry shift towards agentic commerce.
Addressing Trust and Security Concerns
The push for human intervention is largely driven by concerns surrounding trust and security in agentic commerce. Experts like Nikhil Pahwa, founder of MediaNama, emphasize the critical need for guardrails, recourse, and reversibility, noting that a lack of trust can lead to system failures. Pahwa suggests practical solutions such as delegating unique handles to agents (e.g., agent-nixxin@ybl), implementing separate agent PINs that are on by default, and allowing users to set granular transaction limits—per transaction, per month, and per day/week. He also advocates for AI agents to utilize wallets instead of direct bank accounts to mitigate risks. These suggestions highlight the complex challenges of integrating AI into financial systems while maintaining consumer confidence and preventing potential misuse or fraud, especially when contrasted with fully autonomous protocols like Coinbase's x402, which aims for independent, AI-driven commerce without human intervention.
Key points
- MeitY proposes mandatory 'human-in-the-loop' controls for agentic AI payments above specific financial thresholds.
- The proposal is detailed in the Digital Threat Report 2025-26 by CERT-In, MeitY's cyber wing.
- NPCI is reportedly developing a Unified Agent Protocol to enable AI agents to make UPI payments.
- Experts like Nikhil Pahwa advocate for guardrails, recourse, reversibility, delegated agent handles, and user-set transaction limits.
- Companies like Pine Labs have already launched agentic payment protocols, and others are testing AI agents for commerce.
If implemented carefully, these regulations could foster innovation in AI-driven commerce while protecting consumers, leading to highly efficient and personalized payment experiences. The ability to delegate tasks to AI agents could streamline transactions for businesses and individuals, enhancing convenience and productivity within a secure framework.
Without robust implementation, mandatory human interventions could slow down the very efficiency AI agents promise, or worse, fail to prevent sophisticated fraud. The complexity of ensuring recourse and reversibility in India's payment ecosystem, as highlighted by experts, poses a significant risk to consumer trust and financial security.
