Agentic payments could test India’s data-localisation rules: Cashfree’s Reeju Datta

India’s payment rules may need to be reviewed as artificial intelligence agents begin transacting for consumers, particularly because card credentials used by personal agents could be stored outside the country, according to Cashfree Payments cofounder Reeju Datta.

“The rules probably need to be reviewed because this is still very new. Regulators may need to examine the issue,” Datta told ET AI on the sidelines of the Global Fintech Fest 2026.

He said rules on keeping data in India and protecting sensitive financial information were introduced before autonomous agents emerged as a payment option.


The Bengaluru-based fintech expects personal AI agents to increasingly make purchase decisions and payments for users, while merchants will need systems capable of accepting transactions initiated through different agents and protocols, according to Datta. Cashfree Here, one of the company’s offerings, allows payments to be completed inside a chatbot or another conversational interface, he said.

“The original job of a payment gateway was to help people sell on the internet. Today, its job is to help merchants and businesses participate in the emerging agentic economy, where agents make payments on behalf of people or autonomously,” Datta said.

This could create a new challenge for India’s data-localisation framework, he said. An individual could save card information in a vault connected to a personal agent even if the vault sits outside India, allowing the agent to transact with Indian merchants, Datta explained.

“The user may have voluntarily granted that access, but users can act irresponsibly or share card information out of ignorance, leading to a leak. Payment providers need to be wary of that possibility,” he said.

To reduce this risk, Cashfree does not share sensitive payment data with third-party AI models, he said.

“We ensure that sensitive payments-related data is not shared with the model. It remains within our systems and does not leave India,” he said.

Datta said Cashfree also uses role-based access control to limit what an agent can see and do. “The agent does not receive access that the merchant would not ordinarily have,” he said.

An agent working for one merchant can access only that merchant’s information and cannot query another merchant’s data, he added.

Speaking about liability when an AI agent makes a mistake, Dutta said responsibility must remain with the business offering the agent. “The entity offering the agent has to be responsible for its actions, so the agent must be built accordingly,” he said.

The level of autonomy should depend on the action and the merchant’s preferences, Datta said. Merchants may permit agents to contact customers or recover abandoned carts with limited oversight, but they are more likely to require approval when money leaves their accounts, he said.

“Typically, when money is moving out of the system or the merchant is spending money, the merchant may want the agent to seek approval. For example, a merchant may want the system to check before issuing a refund,” Datta said.

Cashfree believes refunds could eventually become autonomous when backed by appropriate guardrails and access controls, he added. However, some boundaries should remain firm. “Sensitive financial data and card information must remain out of bounds,” Datta said.

“The existing principles require data to be localised and sensitive data not to leave the country,” Datta said. “Regulators and policymakers now need to consider how those principles should apply in an agentic world.”