AI Agents Could Soon Make UPI Payments for You: From Groceries to Investments
India is working on a system to identify and monitor AI agents that could eventually make payments on behalf of users, starting with low-value UPI transactions and potentially expanding to investments.
India is working on a system to identify and monitor artificial intelligence agents that could make payments on behalf of users as the country prepares to introduce agentic payments to its digital payments ecosystem.
The proposed system includes a registry designed to identify and track AI agents involved in financial transactions. The registry is expected to form part of the National Payments Corporation of India's planned Unified Agentic Protocol.
The registry is initially expected to focus on AI agents making payments through the Unified Payments Interface (UPI), India's widely used instant payments network. It could later be expanded to other payment methods, including cards and bill payments.
The development comes as NPCI works towards introducing agentic payments on UPI. Under the proposed model, users could give instructions to AI agents, which would then make payments on their behalf without requiring separate approval for every transaction.
The initial applications are expected to involve simple, low-value payments. Routine purchases such as groceries could be among the first use cases, allowing an AI agent to complete several small transactions based on previously provided instructions.
The technology could eventually handle more complex financial decisions. For example, a user could instruct an AI agent to purchase a product once its price drops to a specified level or make an investment when an asset reaches a particular price threshold.
Such capabilities could significantly expand the role of AI agents in everyday financial activity, allowing them to move from assisting users with decisions to executing transactions based on predefined instructions.
However, the wider use of AI agents for payments also raises questions about security, accountability and unauthorised transactions. A registry could help identify and monitor agents operating on the payment network, but it would not necessarily address every risk associated with autonomous financial activity.
Liability is expected to be one of the key issues requiring regulatory clarity. If an AI agent makes an incorrect payment or carries out a transaction that the user did not intend to authorise, rules would need to establish who is responsible.
As AI systems gain greater control over financial transactions, regulations covering authentication, security, user consent and liability are likely to become increasingly important.
India's approach comes amid a broader global effort to develop systems for identifying and authenticating AI agents involved in financial activity. China's Payment & Clearing Association has issued guidelines covering identification, security and risk controls for AI-agent payments.
Technology companies such as Amazon, Google and Microsoft have also developed agent registries as they explore AI-driven commerce and autonomous transactions.
The growing focus on agentic payments reflects the rapid expansion of AI systems capable of acting on behalf of users. While the technology could make routine transactions faster and more convenient, its wider adoption will depend on ensuring that users retain control and have adequate protection against unintended or unauthorised payments.
