NPCI is reportedly developing the Unified Agent Protocol (UAP). This proposed framework is designed to verify and authorize artificial intelligence agents within the Unified Payments Interface (UPI) ecosystem.
Current Capabilities of AI and UPI Limitations
Today, AI agents can perform tasks such as searching, comparing, and recommending options when booking flights or making online purchases. However, the process stops at the payment stage via UPI because only users and authorized payment applications can initiate transactions, not the AI agents themselves.
How UPI Works and the Need for Changes
UPI supports features like AutoPay and Reserve Pay, which allow users to pre-authorize recurring or future payments by setting an spending limit and authenticating them once using a UPI PIN. Nevertheless, this differs from granting AI agents the right to make independent purchasing decisions and conduct payments on behalf of the user. To bridge this gap, NPCI is working on creating UAP.
Functions and Purpose of UAP
According to Business Standard, UAP is designed to create a trusted, common, and interoperable infrastructure. Through this infrastructure, AI agents will be able to register, undergo verification, and receive permission to conduct operations in the UPI ecosystem without requiring changes to the core payment system mechanisms. It is important to note that UAP will not replace UPI; it will function as an additional layer of verification on top of the existing system, ensuring secure interaction between trusted AI agents and UPI while maintaining the current payment infrastructure.
Mechanism of UAP Operation
If UAP is implemented, it will provide the payment system with a way to ensure that the AI agent is acting with the user's consent before processing a transaction. Although details about UAP's operation are currently limited, it is expected to operate on top of the existing UPI infrastructure. The payment flow might look like this: the user gives a command, the AI assistant compares options, selects the best one, and generates a payment request. Then, UAP verifies the registration, trust, and authority of the AI agent to act on the user's behalf. After this, the request is sent through UPI, and the payment is either approved by the user or completed automatically for pre-approved payments within set limits, depending on applicable rules.
Limitations of AI Access to Funds
It is not guaranteed that AI will gain full access to funds. Although the final structure has not been announced, the system is likely to operate similarly to existing UPI features such as AutoPay and Reserve Pay, where users pre-approve spending limits rather than granting unlimited access to their bank account. Business Standard previously reported that the proposed protocol must verify whether the AI agent is authorized to act on the user's behalf, define the scope of this authority, and establish accountability in case these limits are exceeded. However, the ultimate safeguards will depend on how NPCI and the Reserve Bank of India (RBI) develop this framework. Furthermore, NPCI must obtain RBI approval before launching UAP.
Potential Risks and Significance
As payment systems evolve to support AI agents, cybercriminals and fraudsters will also seek ways to exploit them. Granting AI assistants the right to make purchases on behalf of the user creates a new set of challenges in consumer protection and security. Although the technology is in its early stages, it carries new risks, including fake AI agents, unauthorized or erroneous payments, excessive user permissions, and unresolved issues of liability in case of problems. If UAP is implemented, it could place India among the first countries to create a national infrastructure for agent payments, transforming UPI from a manually user-managed system into one that can be utilized by trusted AI agents on their behalf.