THE ESSENTIALS IN 30 SECONDS

  • Defense Minister Shinjiro Koizumi said on 15 November that AI guidelines for ministry staff and SDF personnel will be released at the end of November.
  • They will update the ministry's July 2024 AI basic policy and rest on three pillars, the first being that humans decide on AI use and answer for the results.
  • The text has not been published yet; a separate June 2025 guideline already governs research and development of AI-equipped defence systems.

Japan's Ministry of Defense intends to publish new guidelines on the use of artificial intelligence at the end of November, Defense Minister Shinjiro Koizumi said on Sunday 15 November in Yokosuka, according to a Jiji Press report carried by Nippon.com. The document is aimed at ministry employees and members of the Self-Defense Forces (SDF). It has not been released yet, so its exact wording and scope remain unknown.

What the minister announced

Koizumi described three pillars for the planned guidelines. The first is that people, not machines, make decisions about how AI is used and carry responsibility for the outcomes. The second is to turn data into a tool for national defence. The third is to balance the fast spread of AI with stronger safety measures. He said the government wants to use AI and data to reinforce Japan's defence capabilities and to "show a new direction."

The new text will be built by revising the ministry's AI basic policy, compiled in July 2024. Koizumi acknowledged that the policy is only about two years old but argued that the technology moves quickly enough to justify revisiting it whenever needed.

What already exists

The 2024 policy was the ministry's first on AI. Then-minister Minoru Kihara presented it on 2 July 2024 as a framework for promoting AI use, citing new forms of warfare and Japan's shrinking, ageing population. According to a report at the time, it named seven priority fields, including target detection and identification, command and control, logistics, intelligence analysis, unmanned assets, cybersecurity and administrative efficiency.

A narrower document followed in June 2025: the Responsible AI Application Guideline, which applies to research and development of AI-equipped equipment. According to the ministry's summary, it sorts systems into high and low risk depending on whether AI output affects destructive functions, and requires a legal and policy review confirming compliance with international humanitarian law and that a project is not a fully autonomous lethal weapon operating without human involvement.

Open questions

The upcoming guidelines appear broader than the 2025 equipment rules, covering everyday AI use across the ministry and the SDF. Several points will only become clear once the text is out: whether the human-responsibility principle comes with concrete review procedures, how the data pillar treats classified information, and how the new guidelines will sit alongside the existing R&D guideline. For now, what is confirmed is the timetable and the three principles outlined by the minister, not adopted rules.