Digital Agency Drafts Generative AI Guidelines for Government

The Digital Agency's Draft Guidelines for Generative AI Programs in Government

The Digital Agency in Japan has announced a draft of guidelines outlining the rules for when and how central ministries and agencies can utilize generative AI programs. These guidelines aim to encourage the use of generative AI programs by government entities while ensuring proper regulation for programs deemed high-risk.

The draft stipulates that each government entity should appoint a control administrator responsible for AI use. These administrators must possess a detailed understanding of how AI programs are employed and are required to assess the level of risk associated with their use. This assessment considers factors such as the specific work the program will be used for, whether it will be trained on personal information, and the role of human judgment in evaluating the generated data.

For programs deemed high-risk, the relevant ministry, agency, or other governmental entity must consult with the Digital Agency. The agency will establish an internal advisory body on the safe use of AI programs and will also request providers of generative AI services to verify whether their programs may produce biased outputs.

The Digital Agency's goal is to enable all government entities to integrate generative AI programs into their work processes in accordance with the guidelines. This integration is expected to improve the convenience of services provided to citizens by allowing AI programs to answer questions posed on government websites.