"AI, Buy It for Me"... How Will Card Companies' Roles Change in the Age of AI Payments?

Spread of 'Shopping AI'... Visa and Mastercard Join the Ecosystem

Verification Must Go Beyond Identity to “Who Granted the Authority?”

“Safety Measures Like Revoking Authority and Additional Approvals Are Needed”

As the era of 'agentic commerce'—where artificial intelligence (AI) handles payments on behalf of people—approaches, industry experts predict that the role of card companies will be expanded. In addition to existing identity verification, functions that verify and control the scope of payment authority granted to AI by consumers may become increasingly important. Assigning responsibility for erroneous transactions resulting from AI misjudgment is also emerging as a new challenge.

Reuters Yonhap News

According to sources in the financial sector on September 9, Anthropic, the developer of the AI chatbot Claude, unveiled its vision for building commerce AI agents earlier this month. These shopping agents discern consumer demands, search for products, and assemble shopping carts. However, when it comes to payment, the agent does not process transactions directly, instead connecting with existing payment systems or designated service providers. Anthropic lists Visa and Mastercard among its ecosystem partners.

There are also moves to have AI handle payments directly. In India, a protocol is being prepared to enable AI agents to process transactions on behalf of users through the real-time payment system Unified Payments Interface (UPI). In this setup, AI executes transactions within pre-set conditions and limits, without requiring consumer approval for every payment.

Will Card Companies' 'Trust Management' Role Grow as Autonomous Payments Increase?

In the current payment environment, consumers explicitly indicate their intentions through actions such as tapping a card or entering a PIN. However, in the near future, transactions may occur when consumers pre-set their goals and conditions with an AI agent, even without direct intervention at the time of purchase.

Nevertheless, consumers remain cautious. According to a survey conducted by the global consulting firm Accenture, while 85% of respondents said they were open to searching for products with AI, only 9% were willing to allow AI to purchase items independently. Concerns about privacy and loss of control appear to be significant hurdles. One financial sector official commented, “Just because a payment is approved doesn’t necessarily mean the AI acted in line with the consumer’s intentions. It’s also necessary to verify whether the transaction was executed within the AI’s pre-assigned parameters.”

Industry observers suggest that as the AI agent payment market grows, the role of card companies may further extend to validating the trustworthiness of AI transactions. In addition to identity authentication and detecting suspicious transactions, confirming whether an AI has actually received its authority from the consumer is expected to become increasingly crucial. Choi Seungwoo, a research fellow at KB Management Research Institute, explained, “Rather than focusing on identifying the AI agent itself using traditional financial authentication systems, card companies must verify who granted the authority to the AI.”

Clarifying responsibility when issues arise is also identified as a critical challenge. If an AI makes an erroneous payment or incorrectly interprets a malicious command as a legitimate instruction, determining whether the consumer, the AI development company, or the card company should bear the resulting loss remains ambiguous. Jo Jinhyung, a professor at Gyeongin National University of Education, stated, “To use AI agents safely, users should be able to revoke authority at any time. High-value or atypical transactions should always require human approval.” He further emphasized, “The basis for the AI agent’s decisions and judgments must also be presented transparently.”

This content was produced with the assistance of AI translation services.