India plans AI registry as it looks to roll out agentic payments, sources say

India's payment regulator is constructing a registry to authenticate and oversee AI agents conducting transactions on its domestically developed payments network, according to sources familiar with the matter. The registry will be a crucial component of the National Payments Corporation of India's (NPCI) forthcoming Unified Agentic Protocol, which aims to vet AI agents that handle payments on behalf of users, the sources revealed.

Due to their non-authorization to speak with the media, none of them could be named. The Payments and Clearing Association, which is overseen by the central bank, has released guidelines for AI-agent payments, addressing identification, security, and risk controls. This development by the NPCI aligns with the trend of other governments and payment companies worldwide constructing similar systems to identify and authenticate AI agents.

Amazon, Google, and Microsoft have already developed agent registries for agentic commerce. The NPCI's plan to compile a registry of agents is also motivated by the increasing scrutiny on unethical behavior exhibited by agents at prominent AI companies like OpenAI, Meta, and Anthropic. For instance, AI agents operating on the Unified Payments Interface (UPI) could initially facilitate small, recurring digital payments without necessitating user approval for each transaction, with low-value purchases such as groceries being the initial applications.

The system may ultimately empower agents to carry out more intricate instructions, such as purchasing items when discounts reach a predetermined level or investing when prices cross predefined thresholds, according to the second source. However, the responsibility for resolving issues related to erroneous or unauthorized payments will have to be managed through regulation, as pointed out by the third source.