Bank Salad has developed technology to verify and control errors that may occur during financial transactions involving artificial intelligence (AI) agents.
On October 15, Bank Salad announced that it has filed three patents related to technology that governs the judgment and execution of financial AI agents.
This technology focuses on blocking potential risks that may arise when AI agents utilize financial services on behalf of consumers.
In particular, the company has developed technologies in three areas: judgment of execution targets, numerical verification, and execution control, aimed at preventing the phenomenon of 'hallucination,' where AI generates inaccurate information.
First, the company has created a technology that analyzes credit status during the loan pre-approval process to identify customers who can secure loans at lower interest rates than before. This is achieved by training the AI prediction model with actual loan assessment results to enhance judgment accuracy.
Additionally, Bank Salad has developed technology to verify that errors do not occur when the AI analyzes financial product information and personalized benefits. This involves cross-verifying the figures generated by the AI using financial regulations such as the Debt Service Ratio (DSR) and Loan-to-Value Ratio (LTV).
The technology also includes a feature that suspends the execution of financial transactions that may be unfavorable to customers. Before executing financial services, the AI agent checks the accuracy of the input information and figures, ensuring that transactions do not proceed if errors are detected.
A representative from Bank Salad stated, "Bank Salad considers it a fundamental principle to ensure that financial AI agents provide financial benefits to customers while verifying and controlling errors to enable safe service usage. Through this, we aim to create a reliable financial AI service that customers can use with confidence."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.