Government and Generative AI: Do Cabinet Members Understand Gravity of Responses to Diet Questions?

Editorial

16:17 JST, May 31, 2026

Responses made by the prime minister and cabinet ministers to questions in the Diet articulate how the government will address emerging challenges and help set the nation’s course.

These responses are not merely explanations of policy. They are also part of policymaking and can be described as a creative act. In light of this, the government’s policy to use generative artificial intelligence to draft responses to questions from lawmakers in the Diet is nothing short of astonishing.

In addition to drafting Diet responses, the government intends to utilize generative AI for various other tasks as well. It has launched a major pilot project in which a total of 180,000 employees of central government ministries and agencies are allowed to use Genai (pronounced gennai), an AI platform developed by the Digital Agency. Drafting such responses is reportedly a major task of the AI platform.

For many years, preparing answers to questions in the Diet has been a major factor contributing to long working hours for bureaucrats. This is because they must ensure consistency with an enormous number of past responses by checking laws, regulations, statistics and other information.

It is not objectionable to streamline operations for bureaucrats by entrusting tasks such as gathering data and checking facts to generative AI. It would be meaningful for them to use the time saved to engage in discussing and formulating policies.

However, generative AI involves a system that merely outputs text based on information it has learned from the internet.

In contrast, answers to questions in the Diet must be carefully prepared by not only checking past responses but also considering the people’s lives and comprehensively assessing various other elements related to new issues, such as conflicts of interest and the national good.

Furthermore, past responses in the Diet often contain subtle, unspoken nuances, such as in diplomatic situations, and the same words can take on different meanings depending on the context in which they are used.

For example, then Foreign Minister Masayoshi Ohira made a response in 1972 regarding relations between China and Taiwan, saying: “We hope for a peaceful resolution. We believe there is no possibility of [the two sides] escalating into an armed conflict.”

This statement implies that Japan understands and respects China’s claim that Taiwan is an inalienable part of Chinese territory, but this is contingent on a peaceful resolution.

It is simply impossible to entrust generative AI with making such judgments.

The government has said policy decisions are made by human beings, but if it entrusts generative AI not only with data creation but also with drafting responses to questions in the Diet, it cannot say that such decisions will not be affected by AI.

The burden on bureaucrats regarding Diet responses should first be reduced by lawmakers striving to submit their questions further in advance. If the people giving responses heavily use generative AI, then the people asking questions would also eventually come to rely on the technology. If this happens, it could lead to arguments that ministries and agencies, as well as the Diet, are unnecessary.

(From The Yomiuri Shimbun, May 31, 2026)