Banksalad Files Three Patent Applications to Prevent AI Misjudgments in Financial Transactions

[Financial News] Banksalad has developed technology to control the decisions and execution of artificial intelligence (AI) agents. It is expected to boost confidence in financial transactions by controlling errors that may occur as AI agents provide financial services.
According to industry sources on the 15th, Banksalad has completed filing three patent applications related to “technology for controlling the decisions and execution of financial AI agents.”
The financial sector expects AI agents to evolve beyond simply recommending products to conducting actual financial transactions, such as loan applications and asset management. Accordingly, the importance of control technology that can halt transactions or recheck them when errors occur is expected to grow, alongside the need for accurate AI decisions.
Banksalad focused on building technical safeguards—including determining the target of an action, validating figures, and controlling execution—to prevent “hallucination” errors in which AI generates distorted information and gives false answers.
The technologies covered by the patent applications are designed to block potential risks that may arise when AI agents conduct financial transactions on behalf of consumers. Since even a small calculation error can cause major losses in financial transactions, the focus was on ensuring the stability of AI agents.
One technology selects customers who may qualify for a lower interest rate during preliminary loan screening, taking into account conditions such as their credit status. Training the AI prediction model on actual screening results has improved its accuracy.
Banksalad has also developed technology to check for missing information as AI analyzes financial product details and personalized benefits. It cross-checks, as a final step, figures calculated using formulas required under financial regulations, such as the debt service ratio (DSR) and loan-to-value ratio (LTV).
The technologies also include one that holds off on transactions that would be unfavorable to customers. The key is to verify according to established principles that the figures and information entered are accurate when an AI agent carries out a financial service, and to maintain control through the final stage to prevent incorrect decisions.
A Banksalad representative emphasized, “The principle is to verify and control errors made by financial AI agents so that customers can use the services safely. We will continue to develop highly reliable financial AI services that customers can use with confidence.”

[email protected] Kim Tae-il Reporter