India Wants AI Agents to Make UPI Payments — First It Needs to Know Who Is Acting

India Wants AI Agents to Make UPI Payments — First It Needs to Know Who Is Acting

**NPCI is developing an identity registry for approved payment agents, but authentication solves only one layer; consent, liability and reversibility remain the harder market questions.**

*By PriceVia Fintech Desk | Published September 11, 2026 | Updated September 11, 2026*

Why this matters now

NPCI is developing an identity registry for approved payment agents, but authentication solves only one layer; consent, liability and reversibility remain the harder market questions.

Key points

- NPCI is developing a registry intended to verify AI agents that initiate transactions through UPI on a user’s behalf. - Early use cases are expected to focus on small frequent purchases before moving toward conditional commerce or investing. - The overlooked gap is liability: identifying an agent does not by itself determine who pays when instructions are wrong, compromised or misunderstood.

The numbers

| Metric | Value | Context | |---|---:|---| | UPI operator | NPCI | Registry developer | | Initial use case | Small payments | Groceries and routine buys | | Payment rails | UPI first | Cards and bills may follow | | Human parties | At least 3 | User, merchant, provider | | AI layer | 1 | Authorised agent | | Liability rule | Pending | Still to be defined |

What happened

India’s payments operator is developing a registry that would identify and monitor AI agents allowed to transact over UPI, according to people familiar with the work. The registry is part of an emerging Unified Agentic Protocol intended to support transactions initiated on behalf of users. [S1, S2] Initial applications are expected to involve frequent, low-value purchases such as groceries. More complex conditional buying, investing, cards and bill payments could follow, but the legal and supervisory framework is still being formed. [S1, S3]

What everyone is watching

The first test is consent design: users must know what the agent may buy, from whom, at what price and within what time and spending limits. Revoking authority should be as clear as granting it. The second is dispute handling. Existing payment systems assume a person or merchant initiates an action; agentic commerce adds model providers, software developers and data sources. Fraud and error rules must assign responsibility across that chain.

What the market may be missing

PriceVia analysis: a registry proves identity, not intent. A verified agent can still misunderstand an instruction, use stale prices or optimise for the wrong objective. Permission scopes, transaction logs and explainable approvals will matter more than a simple approved list. Small payments are a sensible sandbox because individual losses are bounded and behaviour is frequent enough to test. Yet large aggregate volumes can still create systemic fraud patterns, so monitoring must work at both transaction and network levels.

Positive case

NPCI creates narrow permissions, strong logs and rapid dispute resolution. Merchants gain low-friction automated demand, users save time and Indian payment infrastructure becomes a standard-setter for agentic commerce.

Downside case

Ambiguous consent, account takeovers or model errors create unauthorised transactions. Fragmented responsibility damages trust, while excessive rules slow adoption or concentrate power among the largest providers.

What would change the story

Watch NPCI specifications, RBI guidance, pilot participants, transaction limits, authentication, liability allocation and refund rules. Public technical standards and a controlled pilot would materially strengthen the story.

Related stocks and themes

NPCI, UPI, banks, payment aggregators, e-commerce, AI agents, digital identity, fraud prevention, consumer protection and fintech regulation.

Reader checklist

Separate the confirmed event from the forward case. Track pilot rules, consent controls and liability framework; then compare those signals with management, regulator or exchange disclosures. The headline establishes why NPCI matters now, but the next measurable milestone decides whether attention becomes durable value. Until that evidence arrives, valuation and scenario claims should remain conditional rather than certain.

PriceVia View

The registry is necessary plumbing, but the decisive innovation is accountable authority. Agentic payments scale only when a user can see, limit, stop and challenge what an AI does with money.

Sources and timestamps

- [S1 — Reuters: registry plan and agentic-payment scope](https://www.reuters.com/world/india/india-plans-ai-registry-it-looks-roll-out-agentic-payments-sources-say-2026-09-10/) — published 2026-09-10; accessed 2026-09-11T10:05:00+05:30 - [S2 — NPCI: UPI product and ecosystem information](https://www.npci.org.in/what-we-do/upi/product-overview) — published accessed 2026-09-11; accessed 2026-09-11T10:05:00+05:30 - [S3 — RBI: payment and settlement systems](https://www.rbi.org.in/scripts/PaymentSystems.aspx) — published accessed 2026-09-11; accessed 2026-09-11T10:05:00+05:30 - [S4 — Global Fintech Fest: programme and announcements](https://globalfintechfest.com/) — published accessed 2026-09-11; accessed 2026-09-11T10:05:00+05:30

Visual disclosure

Hero visual created specifically for this article. Thumbnail text: “WHO AUTHORIZES THE AGENT?”. It is an editorial illustration, not a market-data screenshot.

Market-risk disclaimer

This article is for market education and information only. It is not investment advice, a recommendation or a promise of returns. Prices, regulatory decisions, deal terms and forecasts can change; verify the latest primary disclosures and assess risk independently.