Skip to main content

This page has been translated using TexTra by NICT. Please note that the translation may not be completely accurate.If you find any mistranslations, we appreciate your feedback on the "Request form for improving the automatic translation ".

Guidebook for Countering Risks of Using Text-Generating AI (α Version)

Last Updated:

Overview

In Digital Agency, based on the Priority Plan for the Realization of a Digital Society in (Cabinet decision on June 7, 2022) , we are assessing the actual situation and trends of AI, identifying risks and necessary measures, and considering appropriate use by the public and private sectors.

Recent technological innovations such as generative AI have provided various advantages. The government, while keeping an eye on such technological trends, has repeatedly discussed the business use of generative AI in relevant ministries and agencies at the 10th Meeting for the Promotion of a Digital Society Executive Meeting and the Document Conference. In addition, since December 2023, we have conducted technical verification for the appropriate use of generative AI and published the results *.
*For details of the technical verification results, please refer to the 2023 Technical Review for Appropriate Use of Generative AI in the Digital Agency and Government Administration Conducted .

Based on the history of discussions and the results of verification so far, we will publish it as "Guidebook for Countering Risks of Using Text-Generating AI (α Version)." We have compiled a technical report on risks that can be assumed when actually utilizing generated AI in administrative services, etc. and measures to deal with them.

Publishing Background

When using the generated AI, the Hiroshima AI Process (Ministry of Internal Affairs and Communications) in and the in are available as guidelines for administrative work. Based on these guidelines, the Guidebook for Countering Risks of Using Text-Generating AI (α Version) focuses on the text generation AI and its use among the generated AI, and describes more specific risks and measures to be considered in government procurement.

Characteristics of Guidebook for Countering Risks of Using Text-Generating AI (α Version)

Targeting service developers and providers of text generation AI

Some generated AI can generate not only text but also images, sounds, and videos. Since most of the information exchanged in actual business is text information and the use of generated AI other than text information is in the verification stage in Digital Agency, we focus only on the text generated AI among the generated AI.
In addition, it is assumed that it will be mainly used by persons in charge of system development for utilizing generated AI, and it does not include the risks of infrastructure model developers (risks related to the supply chain of learning data and rights issues, etc.) and risks of service users (risks related to deepfakes, etc.).

Categorization of text generation AI by use form, expected use case, and process

The assumed risks of the text generation AI vary depending on the type of use (embedded in a specific service, web API format for using a large-scale language model, etc.) and use case. In the alpha version, we are aware of the differences in the type of use and use case, and list the risks and points to keep in mind according to the process defined in the Digital Government Promotion Standard Guidelines Practical Guidebook.

Target audience for "Guidebook for Countering Risks of Using Text-Generating AI (α Version)"

We assume a government officials related to the introduction of generative AI into government information systems. Please use it in conjunction with the "Digital Government Promotion Standard Guideline Practical Guidebook" and the "AI Guideline for Business Ver 1.0" .
In addition, if you are considering introducing a text generation AI to improve your company's services or business operations outside of government officials, please use it as a reference to consider possible risks and measures when using the generated AI.

Composition of the Guidebook for Countering Risks of Using Text-Generating AI (α Version)

It consists of six chapters in the first half of this book and three chapters in the second half of this book.

Download Materials

Guidebook for Measures to Reduce Risks during Use of Text Generation AI (α version) (PDF / 1,742 kb) (updated on June 10, 2024)

Update Contents

  • First published on May 29, 2024
  • May 30, 2024 Minor correction of errors
  • June 10, 2024 Minor correction of errors
  • June 6, 2025