Second Advanced AI Utilization Advisory Board

Last Updated:: Feb 4, 2026

Outline

For an overview of the Committee, please see Advanced AI Utilization Advisory Board .

Event Information

Date:: Tuesday, January 13, 2026 from 12:00 p.m. to 2:00 p.m.
Location:: Office of the Prime Minister's Office - Online

Proceedings

1. Opening
2. Proceedings
- 1. Outline of Periodic Report on generative AI System by Each Ministry and Agency
- 2. Initiatives for government AI
- 3. Advanced AI Utilization Advisory Board 1st Working Group Implementation Report
- 4. Trends in generative AI in Japan and other countries
- 5. Draft Revised Guidelines for the Enhancement of the Guidelines concerning the Acquisition and Utilization of generative AI for the Evolution and Reform of Public Administration
3) Closing

Data

Summary of proceedings

At the beginning, Mr. Parliamentary Vice-Minister for Digital Transformation Kawasaki gave presentations on the ideal "reliable AI" that the Japanese government aims to realize, Digital Agency's initiatives related to AI utilization, including governmental AI, and the main topics of the Second Advisory Board Meeting. Next, Mr. Vice-Minister for Digital Transformation, Chief Officer of Digital Agency Misumi expressed his expectation for further promotion of AI utilization within the Japanese government.

1. Outline of Periodic Report on generative AI System by Each Ministry and Agency

The Secretariat made a report in Appendix 1 on the results of an analysis of the state of generative AI utilization by the various ministries and agencies.

The main questions and opinions of the attendees are as follows.

Nabatame: It is important that governance in : As the introduction of generative AI within the Japanese government expands, it is important for the Japanese government to make this initiative a common asset and promote its cross-sectional deployment to all ministries and agencies. The keyword for promoting the utilization is how to standardize it. For example, there may be cases where an initiative has been implemented by one ministry and agency, but has not been introduced by another ministry and agency. It is necessary to raise the level of the efforts of each ministry and agency through support for accelerating the introduction, and to enhance the initiative of the Japanese government as a whole. Therefore, a mechanism that plays a comprehensive role-such as the quarterback function - is necessary.

OKADA Kazuhiro: Regarding the "Examples of Outcome Targets", this is useful information for understanding the needs of the field and the desired direction. At the same time, we believe that a mechanism for sharing best practices that have achieved outcome targets in the GENAI may be effective. We would like to hear the secretariat's thoughts.

Secretariat: The outcome targets have been covered since the Second Periodic Report. Sharing best practices is one of the roles of the Advisory Board. The Advisory Board will continue to share best practices, as achievements will be accumulated through future periodic reports. The significance of sharing best practices through GENAI is as you pointed out, and as described on page 9 of Appendix 2, it has already been partial implementation. In the future, we plan to promote sharing of best practices with various ministries and agencies as reference materials.

2. Initiatives for government AI

The Secretariat reported on the efforts of the Government AI and the outline of generative AI Utilization Circumstance (GENAI) in Appendix 2.

The main questions and opinions of the attendees are as follows.

Yoshinaga: In the graph showing "GENAI Usage Status: Number of Users by Use Case (by App)" on page 11 of Document 2, there is an item called "chat", but if you take "chat" literally, it means chat. generative AI's chat function ranges from simple information search to idea creation and policy making. In the future, what kind of "chat" it will be, and if you break it down, the trend will become clearer.

Secretariat: As described in P. 9, "chat" is defined as "interactive AI", but specific use cases have not been identified. It is possible to make a decision by checking the prompt, but we have not checked it at this point because there is no need for it.

Torisawa: It is difficult for humans to see the prompt and learn how to use it, but it may be possible to automatically learn how to use the prompt by asking the LLM how to use the prompt.

Secretariat: Other ministries and agencies are concerned about the leakage of information, and Digital Agency will consider it to the extent possible.

Nabatame: It is important that governance in Material 1, there are cases where data constraints lead to constraints on AI utilization. By widely sharing what constraints exist at the current stage, we believe that the understanding of data retention and the value of data development will advance.

Secretariat: It is as you pointed out, and we would like to understand where the barriers are in the utilization. From cases where verification was advanced but it did not reach the regular use stage, it is possible to clarify the challenges for the utilization of generative AI. Therefore, we would like to promote the horizontal development of information to each ministry and agency while conducting hearings.

Chairman: Kadobayashi: generative AI, and will Digital Agency conduct hearings?

Secretariat: Prefectural and Municipal Governments, we would like to develop open databases such as white papers to promote GENAI. We would like to create an environment where it is possible to conduct analysis by entering simple questions into the AI. In many cases, domain knowledge held by each ministry and each division and office is not shared even among neighboring departments. While paying attention to permission management, it is necessary to consider how to make such data available to AI agents and others. Since we believe that there are some databases that can be made open to a certain extent within the Cabinet Office and each Ministry, we would like to continue to make efforts.

YUASA Members: issue, will Digital Agency or the relevant ministries and agencies decide how to respond to information disclosure requests from the public regarding "GENAI"? Please tell us the status of consideration at this point.

Secretariat: Basically, it is assumed that each department that holds the data subject to the information disclosure request will respond.

YUASA Members: It is desirable that the various ministries and agencies be able to apply the information disclosure request in a consistent manner. I would like Digital Agency to do its utmost.

There were also opinions that it is important to apply the Common Principles in advance and to update them in an agile manner by reflecting the contents that are considered to be the best at that time, especially for the provision of information on learning data, it was pointed out that it is important to clarify the reference relationship with international documents such as the Hiroshima AI Process Reporting Framework from the viewpoint of ensuring inter-operability. There were also opinions that it is important to apply the Common Principles in advance and to expand the target applications in a phased manner, starting with high-risk applications. There were no opposing opinions regarding the expansion of the target AI of the Guidelines. There were opinions that it is necessary to review and enhance the "Procurement Check Sheet" and "Contract Check Sheet" in light of the latest technological trends and the development of use cases, even if the scope is not yet established. There were opinions that it is important to review and enhance the "Procurement Check Sheet" and "Contract Check Sheet" in light of the latest technological trends and the development of use cases, even if the scope is not yet established. There were opinions that it is important to review and evaluate the so-called "white" efforts, such as voluntary efforts to ensure transparency and reduce risks. Furthermore, regarding the current situation where multiple guidelines are published within the government, there were opinions that it is difficult to refer to them in a cross-sectoral manner, it is difficult to understand the order of precedence, and it is difficult to proceed with on-site implementation. There were opinions that it is necessary to clarify cross-references and to consider integrating them in a phased manner from the viewpoint of a voluntary effort to ensure transparency and to reduce risks. Furthermore, from the viewpoint of reducing the burden on both governments and companies. Regarding the current situation, it is difficult to clarify the cross-sectoral manner, it is difficult to implement in the on-site in a phased manner. There was also pointed out the importance of a mechanism to gradually expand the application of the common principles in december 2025. In light of the guidelines in advance. Regarding the four axes in light of the contents that are considered to be the best at a certain degree of the contents. Furthermore, it is difficult to implement the on-site. Furthermore, the on-site reporting framework such as the on-site reporting framework. In addition, there was pointed out that it is important to review. In addition to the guidelines. In addition to the on-site of a phased manner. In addition to the high-risk management guidelines, and the high-risk. There is necessary to the high-level. Concerning the guidelines in a high-risk guidelines. Furthermore in a cross. Regarding the on-site. For the on-site. As well. Regarding the, while the nature and on-site. In the current. In order to reduce the on-site. In the on-site implementation. In order to reduce the on-site. In order of on-site. In order to reduce the on-site. In the on-site. In order. In order. In In addition to on-site. In addition to reduce, in the current situation in the current in order, and in the on-Guidelines, the on-site, and on-site, and on-the on-site implementation. In the on-site. In order. In order. In the present, and in addition, in addition, in order. In addition to the on-site, and on-site, and on-on-on-on-site. In the on-site. Furthermore, and on-site, on-site implementation on the on-site. the on-site. on-the on-site. the on-site. There. There. In addition. In addition, in addition, in addition in In In addition, in addition, in order. In addition, in addition, in order. in addition, in order in order. In addition, and in addition, in order, in addition, in order of the high-the guidelines in order of the high-In order of the high-in addition, in the onhigh-onhigh-in addition, in addition, in addition, in addition, in addition, in the present On page 10 of Document 2, regarding the dashboard of GENAI utilization status that has been released, it is desirable that there is a monitoring mechanism that can grasp the form in which AI is utilized when deploying to other ministries and agencies in the future. In addition to large-scale demonstration, it is important to introduce a mechanism to verify the effects.

OKADA Kazuhiro: In page 9 of Document 2, there are "general purpose applications" and "AI applications for administrative work." In addition to the common technical and ethical training of the government ministries and agencies, it is necessary for each government ministry and agency to conduct training from an ethical perspective. Human resource development is essential for the safe and secure use of AI. I would like to hear the opinions of the secretariat.

Secretariat: Each ministry and agency will establish a CAIO to strengthen governance. Regarding the guidelines, Digital Agency has provided training materials to each ministry and agency. Regarding training shared by all ministries and agencies, in addition to basic training, training on GENAI is also planned. Regarding GENAI usage, as of August 2025, about half of Digital Agency staff had never entered a prompt. In light of this situation, we believe it is important to first promote utilization by enhancing training for beginners, and then prepare phased training for intermediate and advanced users.

OKADA Kazuhiro: Document 2, "3. Development of Government-wide Database for generative AI," is a very important initiative. On the other hand, in Japan, there is a background where date has been accumulated in the Japanese language. When considering the utilization of date in AI, if it is simply translated into English using the existing generative AI translated function, there is a risk that the delicate nuances of the time may not be expressed appropriately. In particular, regarding the translation of official documents, please tell us if there are any points that are given special consideration.

Secretariat: It is as you pointed out. The final responsibility for the use of the output results will be borne by the staff who use generative AI. However, we would like to call attention to the fact that it is necessary to sufficiently confirm the contents, taking into account the historical background and values of when the official documents were created, when using GENAI.

Chairman: Kadobayashi: : This issue is closely related to R & D. It is desirable that this issue is also discussed at the Information-Technology Promotion Agency, Japan and the AI Safety Institute (AISI), taking into account the trends in other countries.

Chai Shan Member: GENAI, and share them as reminders. If the number of users increases, risk training and literacy improvement efforts will become even more important in the future. At present, expanding the number of users itself is an issue, but it is also necessary to take measures that look ahead to after the expansion.

Secretariat: One of the roles of the Advisory Board is to understand risk case information and give advice on prevention of recurrence. Therefore, if a risk case occurs, we will share it with the Advisory Board and consider measures to prevent recurrence. We will also consider training while paying attention to it.

3. Report of Working Group 1

The Secretariat presented an overview of the first meeting of the Working Group in Appendix 3.

The main questions and opinions of the attendees are as follows.

Yoshinaga: Regarding the "Opinions on Chatbot-Nation Risk" described in "Discussion Point 1: Perspectives of Managing and Reducing Risks When generative AI AI is Widely used by the General Public", we believe that a protocol to guide manned response is extremely important. At the end of 2025, I participated in an external event called the "AI Agent Expo". I found that erroneous answers in AI can be attributed to a lack of communication on the part of humans. Therefore, agent products have emerged to assist in the communication of humans who lack language. For example, a AI agent has emerged that responds to prompts input by humans with only one word, such as "Final Tax Return", by complementing the context while further asking about needs. In the future, there is a possibility that the need for manned response will decrease due to technological development. However, a protocol to guide manned response will be necessary until the technology reaches a certain level. Al Harthi

Torisawa: We fully agree with the issues of communication and hallucination. When classifying hallucination, there are two types of cases: those that present incorrect information and those that are so-called "oversights," in which relevant information exists but is not referred to. Personally, when I needed to report the details related to real estate sales in my past final tax return, I could not understand it because there were many special exceptions, and as a result, I had an experience of asking a certified public tax accountant. Since there are many special exceptions in the tax field, it is necessary to give consideration to the risk that the AI may "overlook" these. In addition, in the case where staff members are manned after communication with the generative AI, the content of the response on the staff side is also considered to change compared to the case where communication with the AI is not mediated. Therefore, it is important to have a system in which information such as issues that users are likely to overlook is shared with the corresponding staff in advance.

Secretariat: Working Group Implementation Report will be shared with National Tax Agency.

There were also opinions that it is important to apply the Common Principles in advance and to update them in an agile manner by reflecting the contents that are considered to be the best at that time, especially for the provision of information on learning data, it was pointed out that it is important to clarify the reference relationship with international documents such as the Hiroshima AI Process Reporting Framework from the viewpoint of ensuring inter-operability. There were also opinions that it is important to apply the Common Principles in advance and to expand the target applications in a phased manner, starting with high-risk applications. There were no opposing opinions regarding the expansion of the target AI of the Guidelines. There were opinions that it is necessary to review and enhance the "Procurement Check Sheet" and "Contract Check Sheet" in light of the latest technological trends and the development of use cases, even if the scope is not yet established. There were opinions that it is important to review and enhance the "Procurement Check Sheet" and "Contract Check Sheet" in light of the latest technological trends and the development of use cases, even if the scope is not yet established. There were opinions that it is important to review and evaluate the so-called "white" efforts, such as voluntary efforts to ensure transparency and reduce risks. Furthermore, regarding the current situation where multiple guidelines are published within the government, there were opinions that it is difficult to refer to them in a cross-sectoral manner, it is difficult to understand the order of precedence, and it is difficult to proceed with on-site implementation. There were opinions that it is necessary to clarify cross-references and to consider integrating them in a phased manner from the viewpoint of a voluntary effort to ensure transparency and to reduce risks. Furthermore, from the viewpoint of reducing the burden on both governments and companies. Regarding the current situation, it is difficult to clarify the cross-sectoral manner, it is difficult to implement in the on-site in a phased manner. There was also pointed out the importance of a mechanism to gradually expand the application of the common principles in december 2025. In light of the guidelines in advance. Regarding the four axes in light of the contents that are considered to be the best at a certain degree of the contents. Furthermore, it is difficult to implement the on-site. Furthermore, the on-site reporting framework such as the on-site reporting framework. In addition, there was pointed out that it is important to review. In addition to the guidelines. In addition to the on-site of a phased manner. In addition to the high-risk management guidelines, and the high-risk. There is necessary to the high-level. Concerning the guidelines in a high-risk guidelines. Furthermore in a cross. Regarding the on-site. For the on-site. As well. Regarding the, while the nature and on-site. In the current. In order to reduce the on-site. In the on-site implementation. In order to reduce the on-site. In order of on-site. In order to reduce the on-site. In the on-site. In order. In order. In In addition to on-site. In addition to reduce, in the current situation in the current in order, and in the on-Guidelines, the on-site, and on-site, and on-the on-site implementation. In the on-site. In order. In order. In the present, and in addition, in addition, in order. In addition to the on-site, and on-site, and on-on-on-on-site. In the on-site. Furthermore, and on-site, on-site implementation on the on-site. the on-site. on-the on-site. the on-site. There. There. In addition. In addition, in addition, in addition in In In addition, in addition, in order. In addition, in addition, in order. in addition, in order in order. In addition, and in addition, in order, in addition, in order of the high-the guidelines in order of the high-In order of the high-in addition, in the onhigh-onhigh-in addition, in addition, in addition, in addition, in addition, in the present In "Discussion Point 1: Perspectives of risk control and risk mitigation for the general public to widely use the generative AI chatbot", areas where the laws and regulations is revised periodically are indicated as future directions.

Secretariat: is a long-term project that is expected to be introduced in 2028. In the future, if the Working Group is consulted depending on the progress, information will be shared at the Advisory Board as appropriate.

4. Trends in generative AI in Japan and other countries

The Secretariat reported the trend of generative AI in Japan and other countries in Appendix 4.

The main questions and opinions of the attendees are as follows.

KITAMURA Tetsuo: Use cases and technological trends in the private sector were introduced. Japan will strengthen strategically weak tasks and further develop strong tasks. It is essential to create new tasks to regain lost ground and to understand the use cases and trends corresponding to them. Therefore, it is necessary to strategically consider the granularity and targets of the survey in the future.

KITAMURA Tetsuo: A list of the current guidelines of each ministry and agency on AI was introduced. However, there are some guidelines that should be translated into English in a timely manner from the viewpoint of enhancing Japan's presence. As our past experience, we translated the AI risk management framework of the U.S. NIST into Japanese in cooperation with relevant US organizations. However, there were many difficulties in the translation process. In the future, as part of an effort by the AI of Government, it is possible that a translated RAG will be established as one of the RAGs in the generative AI User Environment (GENAI). However, at this stage, there is room for discussing the development of a translated dictionary for the guidelines of each ministry and agency on AI in cooperation with relevant organizations.

Secretariat: Guidelines are translated into English, we would like to consider adhering to terms such as "AI Guideline for Business" so as not to cause inconsistency of terms within the Japanese government.

Nabatame: It is important that governance in should be addressed in a cross-governmental manner. It is effective to compare the rules for procuring and using AI adopted by governments in other countries, in terms of both individual efforts by each ministry and agency and cross-governmental efforts, and to consider Japan's measures while paying close attention to the trends of other countries. The joint approach to resolving issues will send an important message about the smooth relationship between Digital Agency and each ministry and agency in the future.

5. Draft Revised Guidelines for the Enhancement of the Guidelines concerning the Acquisition and Utilization of generative AI for the Evolution and Reform of Public Administration

The Secretariat reported on the draft revised policy toward the enhancement of the Guidelines in Appendix 5.

The main questions and opinions of the attendees are as follows.

Chairman: Kadobayashi: , a member of the Committee, will report on the results of the questionnaire survey conducted by Keidanren.

KITAMURA Tetsuo: , we believe that it is necessary to organize the generative AI system, which is highly black-box, in consideration of the balance so that citizens will not be held excessively responsible.

Secretariat: Although each member has reviewed the "Draft of Points to Note When Creating Terms and Conditions for Use of the generative AI System for the Public," the GOJ will continue to pay attention to the creation of the draft so that the contents will not be excessive for the public.

Chai Shan Member: As for the revision of the contract check sheet, we believe that it is an important message based on the trend of the private sector to clearly indicate that the contract is basically a quasi-mandate contract.

Secretariat: We would like to proceed with implementation while taking into account the guidelines of the Japan Deep Learning Association (JDLA) and present a revised draft at the next Advisory Board meeting.

Chai Shan Member: It is understood that the "Reference Sheet for Measures for Intellectual Property Rights, etc." is currently being prepared. While it is a very important effort, it is difficult to completely eliminate risks related to copyrights, etc. It is not realistic to always check whether copyrights are infringed or not for output materials of generative AI. Therefore, rather than placing emphasis on checking whether copyrights are infringed or not, promoting utilization in low-risk fields may be more effective in terms of utilization.

Secretariat: As a supplement, the "Reference Sheet for Measures for Intellectual Property Rights, etc." is positioned as a reference material to clarify the relationship with the existing guidelines and the "Procurement Check Sheet." Important contents shall be reflected in the "Standard Form for Rules for Utilization of generative AI Systems" and the "Procurement Check Sheet." We would like to present the contents to the members at the next Advisory Board meeting.

Chairman: Kadobayashi: GOJ requests USG to take measures while paying close attention to international trends, as there have been cases of large-scale lawsuits related to intellectual property rights, etc. in foreign countries.

Yoshinaga: : I would like to comment on the "Principle Code (tentative name) (Draft) on the Protection of Intellectual Properties and Transparent Utilization towards the Appropriate Utilization, etc. of Cabinet Office," which is currently being examined by generative AI, based on the relationship between IP and "AI check sheets." We recognize that the purpose of the Intellectual Property Promotion Law was to require companies to provide risk information, but the Principle Code requires the disclosure of information such as usage models and learning data. "We are not asking for compulsory disclosure," but the word "comply" in English expression has a very strong nuance. Requiring private companies to disclose their usage models and learning data would be an excessive burden and an obstacle to the "counteroffensive" mentioned in the AI Basic Plan. In the first place, the method of collecting information for creating the AI model also falls under know-how for companies and may fall under corporate secrets, so it is necessary to proceed with the consideration quite carefully. If it is included as requirements for the procurement guidelines, the number of suppliers that can implement it will be limited, and there is a concern that it may hinder the entry of startups, etc., and it is not desirable from the perspective of competition law. In addition, on the governmental side, there is a possibility that GENAI will be subject to information disclosure, which may cause information security risks. It is also necessary to pay attention to the possibility that it may become a target of other countries and increase its vulnerabilities.

Secretariat: Principles Code, the Government of Japan is currently in the process of soliciting public comments.

Chairman: Kadobayashi: - BOM (AI Bill of Materials) section. In particular, the disclosure of information on training data requires careful consideration. In the field of standardization, there is a distinction between Normative References and Informative References. However, sufficient consideration should be given to whether or not to refer to the Principles Code, and if referring to it, whether it should be referred to as Normative or as reference information. We hope that we can devise it from February to March.

Secretariat: It is as you pointed out. This guideline is a normative document, but the content includes informative elements. We would like to proceed with consideration, including making selections according to the content.

Chairman: Kadobayashi: In international documents, there is a distinction between "Annex" and "Appendix". However, in Japanese, the meaning of "Annex" and "Appendix" is not clear. Therefore, at the beginning of the document, it is recommended to specify whether the document is a normative document or a reference document.

KITAMURA Tetsuo: Regarding international discussions on AI BOMs, starting with SBOM, discussions are also underway on the extension to AI systems, so-called AI BOMs. It is necessary to advance discussions in Japan while taking into account such international trends. It is also necessary to give careful consideration to the use of the term "compliance" in relation to the Principles Code. For example, in private sector, there are cases where the term "compliance matrix" is used as a term with a strong meaning close to essential conditions in bidding, and attention is required when using it in policies, etc.

OKADA Kazuhiro: As for the risk assessment logic, we generally support the new logic. On the other hand, from the public's perspective, regarding the deletion of "C. Whether the employee has studied classified information or personal information, etc.," which has been a decision axis in the past, there is a possibility that the public, who do not fully understand the national trend, may feel uneasy if they only see the relevant part. We believe that it would be preferable to add a description in addition to the commentary instead of completely excluding it from the logic.

Secretariat: It is as you pointed out. "C. Whether there is learning, etc. of Information designated" confidential "and" personal "information. We would like to make it easier for readers to understand the relationship between" C. Whether there is learning, etc. of Information designated "confidential" and "personal" information. We would like to make it easier for readers to understand the relationship between "C. Whether there is learning, etc. of Information designated" confidential "and" personal "information. We would like to make it easier for readers to understand the relationship between" C. Whether there is learning, etc. of Information designated "confidential" and "personal" information. We would like to make it easier for readers to understand the relationship between "C. Whether there is learning, generative AI function cybersecurity

Chairman: Kadobayashi: We would like to incorporate the opinions received into the Secretariat. This concludes the exchange of views.