Japan weighs rewrite of its military AI policy

Japanese Defense Minister Shinjiro Koizumi said on Oct. 19 that the basic policy governing how the Ministry of Defense and the Self-Defense Forces use artificial intelligence needs to be reviewed, telling lawmakers that AI development has moved faster than the 2024 guidelines can handle, the Nikkei reported.

“I feel there is a need to review it,” said Defense Minister Shinjiro Koizumi.

Koizumi made the remarks at the House of Councillors Budget Committee in response to a question from Daisaku Hiraki of the Komeito party. He did not give a timeline for revising the policy.

The ministry adopted its first basic policy on promoting the use of AI on July 2, 2024, under then-Defense Minister Minoru Kihara. It identifies seven priority areas: detection and identification of targets, intelligence collection and analysis, command and control, logistics support, unmanned assets, cybersecurity, and more efficient administrative work. Tasks in those areas range from finding and identifying targets in radar and satellite imagery to analyzing large volumes of data gathered from the internet and radio signals.

The policy states that AI is meant to support human decision-making and that human involvement must be ensured when it is used. It also says the ministry will not research, develop or introduce defense equipment whose use is prohibited under international or domestic law.

The ministry has said it is turning to AI partly to reduce the workload on Self-Defense Forces personnel and save manpower as Japan’s birth rate falls and its population shrinks. When the policy was released, the ministry also announced a comprehensive strategy to recruit and train cyber specialists.

In managing the risks of AI, including bias, misuse and malicious use, the ministry has said it draws on the principles in Japan’s AI Guidelines for Business, published in April 2024: a human-centric approach, safety, fairness, privacy protection, security, transparency and accountability.

Koizumi told the committee that any review of the policy would also address the question of leaving decisions to AI on its own. He did not say which parts of the policy would change first.