Japan’s Digital Agency Announces Guidelines on Use of Generative AI; Aims to Improve Web Services of Govt Agency Sites

The Yomiuri Shimbun

16:44 JST, March 29, 2025

The Digital Agency announced Friday a draft of guidelines on rules for when and how central ministries and agencies can use generative AI programs.

An aim of the guidelines is to encourage governmental entities to use generative AI programs if the risks from using it is judged to be low, while it will properly regulate use of AI programs deemed to be high risk.

After publicly collecting opinions on the issue, the agency plans to begin adopting the guidelines in May.

The draft stipulates that officials in the newly established posts of control administrators, who are responsible for AI use, in each governmental entity should have a detailed understanding of how AI programs are used.

These administrators are required to assess the level of risk from several perspectives before using AI programs, including: the details of the work to be done using the program; whether the programs will be trained on personal information; and whether human judgement will be used to assess the generated data.

If the level of risk is judged to be high, the relevant ministry, agency or other governmental entity should consult the Digital Agency, which will create an internal advisory body on the safe use of AI programs.

The agency will also ask providers of the generative AI services to be used by the governmental entities to check in advance whether their AI programs may give biased outputs.

The Digital Agency plans to let all governmental entities introduce generative AI programs to their work processes in line with the guidelines.

In doing so, the agency expects that AI programs will answer questions from users of governmental entities’ websites so that services provided to them can be made more convenient.