Key Points
- Visa has started using authorized AI proxies for real-time payments.
- Banks warn of the risks of errors and unauthorized automatic purchases
- Financial responsibilities in emerging payment systems remain unclear
Artificial intelligence shopping agents are moving from product recommendations to actual financial transactions, which also brings new issues for consumer protection, as payment companies are competing to make autonomous purchasing the mainstream.
On October 20th, Visa announced a real-time pilot project with BILL and Billtrust, allowing certified AI agents to process previously approved corporate invoice payments.
The system uses tokenized payment vouchers and control rules set by the buyer, rather than shared account passwords. However, these agents execute payments that have already been approved by the customer, and do not independently decide what to purchase.
This distinction is very important, as shopping assistants with increasingly powerful capabilities are beginning to make financial decisions.
AI Agents Transition from Recommendations to Real Payments
AI Shopping tools are now capable of comparing products, identifying discounts, and completing certain purchasing tasks with limited human involvement.
The next phase is to allow the software to operate within a preset expenditure limit, and to automatically purchase goods or services when specific conditions are met.
The latest measures taken by Visa indicate that payment infrastructure has begun to support such work processes.
Mastercard has also developed authentication tools for autonomous payments, including its Verifiable Intent framework, which helps to clarify what authorized agents of customers can do.
Meanwhile, Privy, which is under Ripple and Stripe, is developing wallet infrastructure that supports automated cryptocurrency payments.
The application scope of this technology has gone beyond traditional shopping scenarios.
Coinpaper Previous reports stated that the AI agents on XRP Ledger processed 13.5 million payments, yet the total value is still relatively small when summed up.
New issues faced by banks due to unauthorized AI purchases
Independent shopping has brought about problems that traditional payment systems were not designed to address.
Assume a customer authorizes AI to purchase a jacket for $200 on their behalf, but the software selects the wrong size, or in some cases, the purchase ends up costing $400.
Liability may involve customers, merchants, AI providers, or payment processors, depending on the authorized content and the applicable protective measures.
In October, Wells Fargo warned customers that sharing account credentials with the AI tool could expose them to liability due to potential errors.
Industry executives also admit that clear standards of responsibility are still being established.
The Federal Reserve identifies authentication, fraud, and the determination of responsibility for erroneous purchases as major obstacles to broader adoption.
These concerns are similar to those faced by AI in managing personal finances, as automated spending requires reliable safeguards.
Stablecoins have added another layer of payment risk
Autonomous payment may ultimately be achieved through credit cards, bank accounts, and cryptocurrency wallets.
However, the consumer protections provided by these payment methods vary.
Bank card transactions usually operate within existing dispute resolution frameworks, whereas stablecoin payments often lack traditional chargeback mechanisms.
When the software executes incorrect transactions, this distinction becomes particularly important.
The x402 integration of XRP Ledger now supports the use of XRP and RLUSD for automated digital service payments.
Similarly, the new machine payment infrastructure is also expanding the range of financial activities that software can perform.
