Japan's national AI governance system runs on two connected pieces of machinery: a Cabinet body called the AI Strategy Headquarters, and a planning document called the AI Basic Plan. Neither one regulates anyone. Both are created by the AI Promotion Act (Act on Promotion of Research and Development, and Utilization of AI related Technology, Act No. 53 of 2025), and together they are the mechanism through which the Japanese government sets national AI policy direction, coordinates ministries, and decides where to spend attention and money. Understanding what this machinery does, and what it does not do, is a prerequisite for reading anything else in Japan's AI governance landscape correctly.

A Cabinet body created by statute, not a new regulator

The AI Strategy Headquarters (人工知能戦略本部) was established inside the Cabinet on 1 September 2025, under Article 19 of the AI Promotion Act, the day the Act's institutional provisions took effect. The Prime Minister serves as director general (本部長). The Chief Cabinet Secretary and the minister responsible for AI strategy serve as deputy directors general (副本部長). Every other Cabinet minister is a member. This is the same organisational template Japan uses for other cross-government priorities, such as its digital and cybersecurity strategy headquarters: a body with no independent staff or budget of its own, whose authority comes entirely from the fact that every minister who would need to act on its decisions is sitting in the room, or, in practice, signing off on the record.

Day-to-day work is carried out by a secretariat function within the Cabinet Office's Science, Technology and Innovation Promotion Secretariat. The Headquarters has met five times between its inception and July 2026: an inaugural meeting on 12 September 2025, a written circulation meeting (持ち回り開催, a formal decision procedure in which ministers sign off without convening in person) on 21 November 2025, an in-person meeting on 19 December 2025 that settled the draft of the first Basic Plan shortly before its Cabinet decision, a circulation meeting on 19 June 2026, and an in-person meeting on 10 July 2026 immediately ahead of the second plan's Cabinet decision. The pattern is consistent with what the body is for: it convenes formally when there is a plan or a major decision to ratify, and transacts routine business by paper in between.

What the Headquarters actually does

The Act gives the Headquarters one central statutory task under Article 18: draw up the AI Basic Plan, and revise it as necessary. Beyond that, its function is coordination. It can establish specialist investigative councils staffed by outside experts and academics, appointed by the Prime Minister, to prepare draft policy for its consideration. Japan already had a precedent for this kind of expert input: a non-statutory AI Strategy Council (AI戦略会議) met fourteen times between May 2023 and June 2025, feeding advice into the policy debate that eventually produced the Act itself. The statutory Headquarters draws on the same kind of expert process, including an AI expert panel that circulated a public draft for the second Basic Plan's revision in around June 2026, ahead of the Headquarters meeting that finalised the plan in July 2026.

Once a plan is settled, the Headquarters' role becomes monitoring: tracking implementation across ministries, setting benchmarks, and reviewing progress against them. What it does not have is any power to issue binding rules, impose obligations on private businesses, or enforce anything. It sits above the ministries and existing regulators, not beside them as a competing enforcement body. The Personal Information Protection Commission continues to enforce the Act on the Protection of Personal Information. The Financial Services Agency continues to supervise financial institutions using its own instruments. Sector regulators keep their own authority. The Headquarters directs attention and coordinates the government's own agenda; it does not add a new layer of enforceable duty on top of what those regulators already do.

How the Basic Plan gets made and revised

The drafting sequence for the Basic Plan follows a consistent pattern: an expert panel or specialist investigative council prepares a discussion draft, the Headquarters secretariat turns that into a formal draft, the draft goes out for public comment, and the Headquarters then adopts the plan, which is then given effect through a Cabinet decision (閣議決定). The Cabinet Office has indicated the plan is intended to be revised on a roughly annual, sometimes faster, cycle, so that it can keep pace with the technology and international developments rather than being fixed in place for years, as the roughly half-year gap between the first and second plans illustrates.

The first AI Basic Plan (人工知能基本計画) was decided by the Cabinet on 23 December 2025, roughly fourteen weeks after the Headquarters' first meeting. The second plan followed on 14 July 2026, after Headquarters meetings in June and July of that year. Both are issued under Article 18, paragraph 1 of the Act, which requires the government to set out its mid to long term measures for promoting AI research, development and utilisation, consistent with the Act's basic principles.

The first plan: direction, not detail

The first Basic Plan carried the subtitle "Japan's Resurgence through Trustworthy AI" and set the government's headline ambition as making Japan "the country most conducive to developing and using AI in the world." It organised its measures around three cross cutting principles, balancing innovation promotion with risk response, responding to developments in an agile and flexible way, and pursuing domestic and international policy in an integrated fashion, and four substantive pillars: accelerating the social implementation of AI, strategically strengthening Japan's AI development capability, improving the reliability of AI systems, and continuing to transform society through the integration of AI technologies.

Consistent with its function as a strategy document rather than a rulebook, the plan is candid about the problem it is trying to solve. It openly acknowledges that Japan has fallen behind other major economies in AI investment, commercialisation and talent depth, and it identifies the sectors where the government judges Japan already has comparative strength, manufacturing, telecommunications and other infrastructure, the content industry, finance, space, oceans, and agriculture, forestry and fisheries, as the priority areas for AI driven transformation and new business model creation. None of this creates a legal duty on any company in any of those sectors. It signals where ministries are expected to direct programmes, subsidies and coordination effort over the plan period.

The second plan: what changed in half a year

The July 2026 revision kept the same statutory basis and the same basic architecture, but shifted emphasis in several concrete directions that reporting on the revision converged on:

  • Cybersecurity. The revised plan calls for substantially strengthening Japan's capacity to counter AI enabled cyberattacks, citing the growing risk posed by increasingly capable AI systems, and for deepening cooperation with foreign governments and AI developers on this front, including by expanding the capabilities of Japan's AI Safety Institute, which has operated within the Information-technology Promotion Agency, Japan (IPA) since February 2024.
  • Synthetic content and disinformation. It adds support for technology that detects AI generated content, as a countermeasure to AI enabled misinformation and disinformation.
  • AI sovereignty and agentic AI. The plan calls for accelerating deployment of agentic AI technologies while pursuing an open approach to AI sovereignty, avoiding excessive reliance on any single vendor or country, and encouraging domestic AI development aimed at problems specific to Japan.
  • International standing. It positions Japan to play a larger role in international AI governance, building on the diplomatic groundwork Japan laid through the G7 Hiroshima AI Process.
  • A commitment to keep reviewing the legal framework. Several outlets reported that the revised plan adds language committing the government to proactively and continuously review Japan's AI related legal system, including the AI Promotion Act itself, so that it can respond to risks more effectively. That has been read, reasonably, as leaving open the possibility that Japan strengthens the Act in future, potentially including penalties for conduct that infringes on people's rights. It is important to be precise about what this is: a plan level commitment to keep reviewing the law, not a change to the law. The Act remains, as of this revision, a promotion statute with no penalties, no prohibitions and no mandatory conformity assessment.

Read together, the two plans show the Basic Plan doing exactly what a national strategy document is designed to do: the first plan set direction and named priority sectors, and the second plan adjusted that direction in response to events, most visibly the cybersecurity and disinformation risks associated with more capable AI systems, without altering the underlying legal status of anything.

What this means in practice for organisations

For any organisation trying to work out what it actually has to do because of Japanese AI policy, the Headquarters and the Basic Plan sit outside the compliance perimeter. They are planning and coordination instruments. They tell you what the Japanese government intends to prioritise, fund, and eventually might regulate, not what you are currently required to do. Three practical conclusions follow.

First, nothing in either Basic Plan creates an enforceable obligation. The plans do not impose registration, disclosure, testing or conformity assessment requirements on businesses. The AI Promotion Act itself, which the plans implement, remains a framework law whose only business facing duty is the non binding Article 7 expectation to endeavour to cooperate with government AI measures.

Second, what actually binds organisations using or developing AI in Japan continues to come from existing law applied to AI, not from this new AI specific layer. The Act on the Protection of Personal Information, enforced by the Personal Information Protection Commission, governs personal data used in or produced by AI systems. Article 30-4 of the Copyright Act governs the use of copyrighted works for AI training. Sector regulators, including the Financial Services Agency for financial services, retain their own supervisory authority, and organisations in regulated sectors such as healthcare or manufacturing should confirm their obligations against the relevant sector regulator directly rather than treating the Basic Plan as a source of sector rules.

Third, the Basic Plan is worth reading as an early warning system rather than a rulebook. Its priority sectors indicate where government funding, pilot programmes and regulatory attention will concentrate next. Its stated intent to keep reviewing the AI Promotion Act, repeated and sharpened in the July 2026 revision, is the clearest signal available that Japan's currently penalty free AI law could gain enforcement teeth in a future revision. Organisations operating in Japan's AI sector should treat the AI Guidelines for Business, the joint METI and MIC voluntary framework now at version 1.2, as their practical standard of care in the interim, and watch each Basic Plan revision for signs of where that standard, or the Act underneath it, might eventually harden into something enforceable.

Primary sources

Related articles