India’s National Payments Corporation of India is developing a registry to verify and monitor artificial intelligence agents that make payments over the Unified Payments Interface, as the country’s real-time payments network prepares for agentic transactions.
The registry would form part of NPCI’s planned Unified Agentic Protocol and initially focus on agents operating through UPI, Reuters reported, citing people familiar with the work.
The idea addresses a problem that is becoming more important as AI systems move from recommending products to initiating transactions: a payment network needs to know which agent is acting, who authorized it and how its activity should be monitored.
Registry would start with small, frequent transactions
NPCI is expected to begin with relatively small and frequent payments such as grocery purchases. More complex use cases, including conditional purchases or investment-related actions, could follow later as the framework matures.
The registry could also eventually extend beyond UPI to payment cards and bill-payment systems, according to Reuters.
UPI is already one of the world’s largest retail fast-payment systems by transaction volume, which makes India an unusually large testing ground for agentic payments. The network processed 24.51 billion transactions worth RMB29.82 trillion rupees in August, according to separate NPCI figures cited by Reuters this week.
An agent registry would not itself resolve every risk. It can create a record of approved software identities and their operators, but questions remain around consent, transaction limits, errors, unauthorized purchases and who is liable when an agent behaves outside a user’s instructions.
Payment networks are converging on the identity problem
The Indian initiative arrives as other payment ecosystems develop their own approaches to identifying AI agents. TNGlobal reported this week that Ant International, Mastercard and Visa are working on a Know-Your-Agent interoperability framework intended to let networks recognize common trust signals while retaining their own verification and decision-making processes.
The approaches are not identical. NPCI’s proposed registry is tied to a national payments infrastructure, while the Ant, Mastercard and Visa work is aimed at interoperability across commercial networks, wallets and agent platforms. Both, however, reflect the same emerging requirement: software that can spend money on behalf of a user needs an identity and accountability layer of its own.
China’s Payment & Clearing Association has also issued guidelines around AI-agent payments, while technology companies including Amazon, Google and Microsoft have been developing ways to register or identify agents in broader software ecosystems.
Regulation still has to catch up with agent autonomy
The registry would create technical infrastructure, but formal rules around liability and user protection are still developing.
A straightforward card or UPI payment normally has a clear payer, merchant and transaction record. An agent can complicate that chain because it may interpret a user’s broad instruction, choose among products, decide when a condition has been met and execute the transaction without another human confirmation at the final step.
That makes controls such as spending limits, explicit scopes, revocation, audit trails and clear operator identity important even before highly autonomous purchasing becomes common.
NPCI’s work suggests India is trying to establish those foundations while agentic payments are still early. Given UPI’s scale, the design choices could influence how banks, wallets, merchants and AI developers elsewhere think about agent identity and transaction accountability.
Ant International, Mastercard and Visa develop KYA framework