Kanazawa, Japan, April 16, 2026 – (JCN Newswire) – Mitani Sangyo Co., Ltd. (hereinafter referred to as Mitani Sangyo), a general trading company and manufacturer founded in Ishikawa Prefecture in 1928, today announced that it has developed an “AI Reliability Governance Framework” that manages the entire response process from AI input to output. Mitani Sangyo has also filed a provisional patent application in the United States for the seven elemental technologies that make up this framework.
This framework focuses on ensuring verifiable sources of information and allowing AI systems to stop generating answers when uncertainty is high.
This effort enables systems to provide evidence-based responses and refrain from responding when uncertainty is high, increasing the reliability of AI-generated responses and addressing persistent challenges in the rapidly advancing business use of generative AI. Mitani Sangyo positions this framework as an AI governance technology originating from Japan, and is working to build intellectual property that will contribute to global discussions on AI safety.
why is it important
Generative AI has rapidly advanced with improved inference models, search augmented generation (RAG), and advances in AI agents. However, even with these systems, the risk of plausible but erroneous responses, so-called hallucinations, remains a challenge, especially in mission-critical business environments.
When leveraging AI for critical business tasks, it’s essential to not only get the right answers, but also to ensure verifiability and trustworthiness. For example, you can check the basis of the information or stop responding if you are unsure. In fields such as finance, law, manufacturing, and public services, where extreme accuracy and explainability are required, preventing erroneous output is a priority over conversational fluency. This requires explainable features such as verification of data provenance and information time frame (freshness), as well as fail-safe mechanisms to detect uncertain responses and preserve them against human intervention.
To address these challenges, Mitani Sangyo’s AI Reliability Governance Framework provides step-by-step controls through three checkpoints: input, process, and output that control how AI systems generate responses and ensure that information is traceable, consistent, and secure before a response is provided.
Overall picture of AI reliability governance framework and value provided
The AI trust governance framework introduces three checkpoints (input, process, output) along the data flow (input -> process -> output), each designed to reduce the chance of illusions and increase traceability.
At input checkpoints, reference data is tagged with evidence such as when and where it was captured, allowing the system to distinguish between information that should be allowed and what information should be filtered out.
Process checkpoints combine multiple technologies to improve accuracy and speed, including a mechanism to determine whether the knowledge possessed by AI is reliable, information compression technology to efficiently organize data, a determination method based on the range of reliability, and technology to quickly determine the boundary between positive and negative results.
At the output checkpoint, the framework evaluates the reliability of the answer and the associated risks and switches its behavior accordingly. Return a response, hold it, or escalate it to a human review.
Together, these controls ensure that all responses are supported by traceable, up-to-date, and consistent information, systematically ensuring the reliability required for mission-critical operations.
Comments from the inventor
“The important thing is not how well AI can answer the question, but whether it knows when to stop,” said Hiroaki Nakano, Information Systems Division, Mitani Sangyo Co., Ltd. The biggest risk in real business is not that AI will fail silently, but that it will confidently answer when it shouldn’t.
This concept didn’t come out of a large lab. This was born out of a practical implementation where we had to repeatedly ask when to trust AI and when to retreat from it.
To use AI in mission-critical operations, organizations must be able to explain when the underlying information is obtained and where it comes from. Additionally, the system itself must be able to avoid situations where uncertainty is high. Through this framework, we transform uncertainty from a hidden risk to something that can be explicitly managed, enabling AI systems that provide not only answers but also non-answers that are justified with transparency and accountability.
Future roadmap
Mitani Sangyo will proceed with PCT application based on US provisional patent application* Going forward, we will promote the practical application of a reliable platform for the use of AI in important businesses through joint demonstrations and partner collaborations with companies and research institutions in Japan and overseas. At the same time, Mitani Sangyo will continue to share technical knowledge through academic publications, providing rationale and ensuring transparency.
Mitani Sangyo will continue to develop AI frameworks that are both easy to use and safe. Mitani Sangyo hopes that in the future companies will be able to introduce and utilize highly reliable AI with confidence.
*The PCT (Patent Cooperation Treaty) is an international patent system that allows a single application to seek protection in multiple member states.
About Mitani Sangyo Group
Mitani Sangyo is a general trading company that was established 98 years ago in Kanazawa City, Ishikawa Prefecture, Japan, and 31 years ago in Vietnam. The company operates in six segments: chemicals, information systems, plastics/electronics, air conditioning equipment, housing equipment, and energy, mainly in the Hokuriku region, the metropolitan area, and Vietnam. Mitani Sangyo is not only a trading company, but also a manufacturer and consultant, pursuing optimal solutions for our customers, and continuing to evolve as a venture company with nearly 100 years of history. Fiscal year ending March 2025: Consolidated sales are 103,072 million yen, and the number of consolidated employees is 3,563.
Please visit our website for more information.
https://www.mitani.co.jp/JA
Media and business contacts
Media inquiries:
Mitani Sangyo Co., Ltd. Public Relations Planning Office (within Kyodo Public Relations)
(In charge: Takizawa, Yu, Kishimoto)
Email: mitani-pr@kyodo-pr.co.jp
Inquiries regarding corporate customers and technology/research partnerships:
Mitani Sangyo Co., Ltd. Information Systems Division
https://forms.office.com/r/0PZrTM8Mng (link to web form)
(In charge: Nakano, Doi)
Press release: https://www.acnnewswire.com/docs/files/20260416_EN.pdf
Topic: Press Release Overview
Source: Mitani Sangyo Co., Ltd.
Sector: Enterprise IT, Digitalization, Artificial Intel [AI]
https://www.acnnewswire.com
From Asian Corporate News Network
Copyright © 2026 ACN Newswire. Unauthorized reproduction is prohibited. A division of Asia Corporate News Network.
