What Is the Draft Second AI Basic Plan? Japan AX, a Stronger AISI and the Impact on Corporate Practice

On July 10, 2026, the government held the 5th meeting of the AI Strategy Headquarters (Japanese) at the Prime Minister's Office and adopted the draft of the second "AI Basic Plan." This is the first revision of the AI Basic Plan under the AI Act. Under the banner of "Japan AX," it sets out public and private investment in vertical AI and physical AI, an evaluation framework for high-performance AI, and stronger functions for the AISI (Japan AI Safety Institute). In this article, I confirm the content adopted and its legal nature, and then explain what companies that develop, provide or use AI services should check at this point. Note that what can be confirmed as of the time of writing (July 12, 2026) is the "draft" adopted by the AI Strategy Headquarters, and a final version approved by the Cabinet has not been published.

What Was Adopted at the 5th Meeting of the AI Strategy Headquarters

The agenda of the 5th meeting of the AI Strategy Headquarters consisted of two items: the "Basic Plan under the AI Act (Draft)" and the "Interim Summary of the Sector-Specific Strategies for Vertical AI" (Cabinet Office, materials for the 5th meeting (Japanese)). The adopted AI Basic Plan (Draft) (Japanese) is subtitled "Japan AX: Stronger and More Prosperous," and the materials also refer to it as "Phase II." At the preliminary draft stage, a call for public comments (Japanese) was conducted from June 19 to 23, 2026.

The draft plan starts from the recognition that agentic AI has grown rapidly and has advanced from a tool that supports work to an actor that makes and executes decisions. On that basis it sets out four principles, "balancing the promotion of innovation and the response to risks," "challenge and learning," "agile response" and "integrated domestic and international policy promotion," and four basic policies corresponding to "using AI," "creating AI," "enhancing the reliability of AI" and "collaborating with AI." As Japan's path to winning, it focuses on sector-specific vertical AI and physical AI that operates in real space, and states that it will advance "AI transformation (AX)," in which society as a whole reviews decision-making and ways of working on the premise of AI. The establishment of "AI sovereignty," avoiding excessive dependence on specific countries or companies, and a policy of revising the plan itself "every year for the time being" were also expressly stated. In a footnote, the draft plan assumes public and private investment in vertical AI, physical AI (especially AI robots) and semiconductors of JPY 23.1 trillion, JPY 10.5 trillion and JPY 68.0 trillion, respectively, by fiscal 2040.

The Interim Summary of the Sector-Specific Strategies for Vertical AI (Japanese), discussed at the same time, sets out a total of 20 sectors in three categories as priority support areas: marketability (manufacturing, shipbuilding, logistics and transportation, information and communications, finance, drug discovery), public interest (medical care, nursing care and welfare; agriculture, forestry and fisheries; construction; education; public administration; energy) and strategic importance (defense, police, disaster prevention, firefighting, cyber, maritime, space, scientific research), and states that a roadmap through 2030 will be prepared. At the meeting, Prime Minister Takaichi stated that public and private investment in vertical AI and physical AI would be promoted using the "Strong and Prosperous Japan" investment framework, and that the functions and structure of the AISI would be strengthened.

Impact of the Draft Second AI Basic Plan on Corporate Practice

As a premise, let me confirm the legal nature of the Basic Plan. The Basic Plan is a plan established by the government under Article 18, paragraph 1 of the Act on the Promotion of Research, Development and Utilization of AI-Related Technologies (Act No. 53 of 2025; the "AI Act") (Japanese), and it is addressed to the government. The AI Act itself is not a regulatory statute that imposes obligations backed by penalties on businesses, and the responsibilities of utilizing businesses under Article 7 of the Act consist of endeavoring to use AI in accordance with the basic principles and cooperating with the measures of the national and local governments. Accordingly, I think the adoption of this draft does not immediately give rise to any new legal obligations for companies.

That said, the draft plan contains several statements that may have ripple effects on corporate practice.

On the institutional side, the draft expressly states that research and other activities under Article 16 of the AI Act will be operated more flexibly and that "systems, including the AI Act, will be reviewed proactively and continuously." It also sets out the development of various guidelines, including a legal analysis of infringements of rights associated with the use of AI, and the building of a mechanism for gathering technology trends, risk incidents and evaluation findings at the AI Strategy Headquarters and prompting each ministry and agency to check and review its systems and guidelines, on the premise that AI-related rules will move on a yearly basis.

On the evaluation framework, the draft states that, centered on the AISI, it will establish and strengthen technical evaluation capabilities enabling AI model evaluation, traceability, technical controls such as guardrails, and information sharing in a crisis, and will put in place a mechanism connecting evaluation results to improvements in systems, guidelines and operations. It also expressly mentions the formulation of guidelines to ensure safety and security in the use and development of AI, including agentic AI; the active evaluation of high-performance AI from the perspective of deterring events that threaten national security; and the use of the international network of AISIs, including the U.S. CAISI and the UK AISI. If evaluation results are connected to government procurement and guidelines, they may become a de facto reference standard in private companies' model selection and vendor review as well.

On data and personal information, the draft refers to the amendment bill to the Act on the Protection of Personal Information, which includes an exception making the data subject's consent unnecessary for providing personal data to third parties where it is used solely for AI development and the like that can be characterized as the preparation of statistics and the like, and the creation of an administrative monetary penalty system, and states that "if enacted," subordinate regulations and the like will be developed toward smooth enforcement. This amendment bill was passed at a plenary session of the House of Councillors on July 10, 2026, the same day the draft plan was adopted, and has been enacted. The draft plan's wording reflects the time before enactment, and the subject of future monitoring has shifted from Diet deliberations to promulgation and the development of the Cabinet Order specifying the effective date and subordinate regulations.

On intellectual property and liability, the draft lists promoting the return of value to content holders; preparing guidelines on the interpretation of tort law and the like regarding infringement of publicity rights and other personality rights through unauthorized use of voices, and considering the need for legislation including an amendment to the Unfair Competition Prevention Act; considering the appropriate approach to intellectual property rights concerning AI-generated output; and continuing to consider the allocation of responsibility when infringement of rights or damage occurs. All of these are measures at the consideration stage, and they do not change the interpretation or application of current law.

Relationship Among the AI Act, the Basic Plan and the Sector-Specific Strategies

The AI Act is a promotion act promulgated on June 4, 2025, and it sets out basic principles, the responsibilities of the national government, local governments, research and development institutions, utilizing businesses and others, the formulation of the Basic Plan, the establishment of the AI Strategy Headquarters and other matters. The framework of the Act and its impact on businesses are covered in Practical Issues in Compliance with the AI Act and the AI Guidelines for Business. As for the Basic Plan, Article 18 of the AI Act requires Cabinet approval and publication, and the same procedure applies mutatis mutandis to amendments. The first AI Basic Plan was approved by the Cabinet on December 23, 2025 (Cabinet Office, AI Basic Plan (Japanese)), and the draft second plan will likewise be finalized as a government plan only upon Cabinet approval.

Positioned under the plan are the "Guidelines for Ensuring the Appropriateness of Research, Development and Utilization of AI-Related Technologies," adopted by the AI Strategy Headquarters on December 19, 2025, and various guidelines consistent with their purpose. The Sector-Specific Strategies for Vertical AI are an execution document that makes the priority investments set out in the plan concrete for each sector, and the current one is, as its name indicates, at an interim stage.

From the perspective of rules that actually apply to companies, none of the documents in this hierarchy is a direct norm of conduct; what binds companies are existing laws such as the Act on the Protection of Personal Information, the Copyright Act and the Unfair Competition Prevention Act. I think it is appropriate to read the Basic Plan as a document that shows the direction in which these laws and related guidelines will be reviewed.

What Companies Should Check Now

In AI procurement and vendor review, one option is to prepare a slot in the review items so that AISI evaluations and guidelines can be referred to once they are published. Specifically, if you add fields to the vendor review form for whether public evaluation or safety information exists for the model used and how the vendor explains technical controls such as guardrails and logs, you can quickly incorporate them into operations after publication. The contractual items to check are explained in AI Vendor Due Diligence and Contract Review. In my view, companies that provide AI services to the national or local governments should add the concrete development of procurement requirements to the items they continue to monitor, because the draft plan sets out "appropriate procurement" and leading adoption by the government itself.

For AI use rules, a useful reference is that the draft plan, on the premise of agentic AI, notes approaches to human involvement such as human in the loop, human on the loop and human in the lead. Setting out in internal rules what scope of execution authority agentic AI is given for which tasks, and who checks or stops it and at what point, can help ensure explainability in the event of an incident, even before the legal analysis of the allocation of responsibility is settled. For rule design, see How to Create a Generative AI Use Policy.

As for the board of directors and senior management, the draft plan states, regarding the risk of cyberattacks that misuse AI, that "risk countermeasures will be carried out organization-wide under the leadership of senior management." One option is to specify, within the system for reporting to the board, the department responsible for tracking the finalization of the Basic Plan and system reviews and the frequency of reporting, together with the company's AI use status. The items to check are organized in A Checklist for Discussing AI Governance at the Board of Directors.

On the other hand, the draft plan itself does not give rise to any obligation to immediately revise contracts or internal rules. The enactment and enforcement of the amendment to the Act on the Protection of Personal Information, the AISI's guidelines, and the legal analysis of the allocation of responsibility and intellectual property are items to be placed on a tracking list so that, once published, the company can decide whether they need to be reflected in its rules and contracts.

Government Materials to Watch and Concrete Development of the System

As of the date of writing, the following materials and developments are unpublished or undecided, and they need to be monitored on an ongoing basis.

• Cabinet approval of the second AI Basic Plan and publication of the final version and the results of the call for public comments (Cabinet Office, AI Basic Plan page (Japanese))
• The final summary of the Sector-Specific Strategies for Vertical AI and the roadmap through 2030
• Publication by the AISI of guidelines on the safety and security of agentic AI and of evaluation information
• Promulgation of the amendment to the Act on the Protection of Personal Information enacted on July 10, 2026, and the status of development of the Cabinet Order specifying the effective date and subordinate regulations
• The status of preparation of guidelines on unauthorized use of voices and publicity rights, and of consideration of an amendment to the Unfair Competition Prevention Act
• The status of research under Article 16 of the AI Act and of the development of various guidelines, including the legal analysis of infringements of rights

All of these are organized on the premise of the "draft" plan as of the time of writing. Because the content may be modified before Cabinet approval, the starting point for a response is to check the differences from the items listed in this article once the final version is published.