ITOCHU Techno-Solutions Corporation Integrates Dynamo AI’s Guardrails to Enhance AI Governance Capabilities in "C-NOAH", Cloud Service for Financial Institutions

ITOCHU Techno-Solutions Corporation (headquartered in Minato-ku, Tokyo; Tatsushi Shingu, President and CEO; hereinafter "CTC") has incorporated Dynamo AI’s governance capabilities into "C-NOAH," our service, for financial institutions to enable secure and compliant use of a suite of applications, system platforms and AI services. By leveraging the AI guardrailing and auditing tool "DynamoGuard" from Dynamo AI (headquartered in California, USA; Vaikkunth Mugunthan, CEO), this enhancement aims to suppress hallucinations and block noncompliant AI use to improve the reliability of C-NOAH's AI chatbot.

In recent years, more and more companies are utilizing generative AI to advance and streamline their business operations. However, many are running into the issue of “hallucinations”, where AI can generate responses containing false information that are not grounded in facts. Generative AI can generate hallucinations based on misinterpreting prompts or insufficient quality and quantity of training data. Companies are taking countermeasures such as educating employees on prompt creation, regularly reviewing training data, and limiting the scope of AI use in business operations.

By leveraging DynamoGuard, CTC has incorporated the capability to suppress hallucinations within C-NOAH's generative AI chatbot. In financial institutions, providing false information could lead to customers or employees making incorrect investment decisions, or leakage of customer information. The newly added function is expected to improve the accuracy of our AI chatbot, thereby enhancing the reliability of responses to inquiries in internet banking and FAQ systems of financial institutions.

DynamoGuard is an AI auditing tool that verifies the consistency between inputted prompts and corporate compliance policies in real time. If an AI prompt or response is deemed inappropriate based on custom compliance policies created in the DynamoGuard platform, the prompt is immediately blocked or flagged by DynamoGuard, and response generation is automatically stopped. Additionally, generated responses are checked against highly reliable information such as official data from public sector agencies, industry associations and corporations, and their accuracy is quantified numerically and assessed. If the assessment results fall below a certain threshold, an alert is displayed on the screen to notify users.

Going forward, CTC will remain committed to strengthening AI governance with measures including hallucination suppression and real-time compliance checks in DynamoGuard, while continuing to customize C-NOAH and develop new applications tailored to customers' business needs. With these efforts, we will support the advancement and efficiency improvement of financial industry business operations through further enhancement of generative AI reliability and safe utilization.