While an era where artificial intelligence (AI) handles everything from product searches to purchases and payments is emerging, there are still many restrictions in financial transactions using bank accounts. As global tech giants accelerate the expansion of AI payment services, banks are taking a cautious stance regarding external AI access to accounts due to security concerns and issues of responsibility in case of incidents.
According to the financial sector on the 18th, 'agent commerce,' where AI searches for products and completes purchases and payments on behalf of consumers, is becoming more prevalent. In fact, a reporter used the AI agent 'Pickle' to purchase sneakers worth 169,000 won (KRW), with the AI handling everything from product search and size selection to connecting to the Naver Pay payment window. The only step the user had to complete was entering the final payment password.
The global payment industry is also moving to expand AI payment services. OpenAI launched the 'ChatGPT Wallet' on the 13th in collaboration with Visa, Mastercard, American Express, and Chase. This service, which is being introduced to some users in the U.S., allows transactions to proceed using a payment token instead of a card number once the user approves the purchase.
In contrast, access to financial transactions such as bank account inquiries or remittances by external AI remains limited. Currently, there is no system in place in the banking sector to identify whether the entity requesting a transaction is an AI agent or the actual account holder. The authentication system in the financial sector is also designed around humans, requiring additional verification such as passwords or biometric authentication for transactions involving the movement of funds, like remittances.
If AI establishes itself as a new gateway for financial transactions, changes in the interaction between banks and customers are expected. This could lead to a scenario where consumers compare financial products or execute transactions on AI platforms without directly accessing bank apps. However, banks maintain that it is challenging to grant external AI access to their systems due to the complexity of products and transaction procedures, such as savings accounts, loans, and foreign exchange.
A particular concern is the issue of responsibility in the event of an incident. While card payments have procedures for cancellations or dispute resolutions, recovering funds from account transfers is not easy since the money moves directly. If AI misidentifies the transfer amount or recipient, it could lead to financial losses, but it remains unclear how much responsibility the bank should bear in such cases.
An official from a major bank stated, "If a transaction is entrusted to an AI agent, it can be argued that the user is responsible in principle, but if an incident occurs, customers may raise the issue with the financial company, questioning whether the bank should have prevented it in advance." He added, "In a situation where it is difficult for banks to fully control the operation of external AI, they have no choice but to be cautious about taking on responsibility for incidents."
Foreign banks are also wary of external AI access. JPMorgan Chase in the U.S. prohibited external AI from accessing customer accounts using login information as of the 27th of last month. However, access through approved application programming interfaces (APIs) by the bank is permitted.
Financial authorities are also working on regulatory improvements. The Financial Services Commission announced in August that it would promote the introduction of 'My AI Agent,' which allows AI to apply for and execute financial services within the scope permitted by the user. Future tasks include establishing an authentication system to verify the identity and transaction authority of AI and setting standards for sharing responsibility in case of incidents.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.