Clifford Chance

Countries around the world are taking diverging approaches when it comes to regulating artificial intelligence (AI), and in particular with respect to generative AI. Some jurisdictions are beginning to regulate, including in the EU and China, while others issue guidance frameworks and rely on existing laws such as data privacy and intellectual property laws to govern the use and deployment of AI tools. Japan has, to date, taken the latter approach, issuing a number of AI-specific governmental guidelines and reports . Now, a comprehensive set of non-binding guidelines have been issued, which will be an important reference for those who run businesses in Japan.

The government has produced a useful, detailed summary of the guidelines (Main body / Annex).  In this article we provide a quick snapshot of the guidelines as well as some takeaway suggestions for commercial organisations to consider.

Background

On 19 April 2024, Japan's Ministry of Internal Affairs and Communications (MIAC) and Ministry of Economy, Trade and Industry (METI) published the AI Operator Guidelines marking a significant step in the country's ongoing engagement with  AI regulation. The latest Guidelines aim to consolidate three key previous publications: the AI R&D Guidelines, AI Utilization Guidelines, and Governance Guidelines for the Implementation of AI Principles. Reflecting the dynamic nature of international discourse, these Guidelines offer user-friendly advice to support AI operators and are designed as a 'living document', expected to adapt and evolve over time.

Structure and Coverage

The Guidelines articulate principles and actions for different players across the development chain and lifecycle of an AI system, namely AI Developers, AI Providers, and AI Users. The Guidelines apply within a commercial or business context, explicitly excluding non-commercial uses, data providers, and indirect beneficiaries. An 'AI system' is defined broadly as any system, including software, capable of autonomous operation and learning, such as machines, robots, and cloud systems. For 'Advanced AI Systems', which include generative AI systems, the Guidelines provide additional principles.

Comprising a main body and an appendix, the Guidelines first outline the underlying philosophies - 'why' principles are necessary and 'what' they are—before the appendix offers practical advice as to 'how' to implement the principles. The overarching philosophy promotes AI as a public asset, contributing to global sustainability and societal transformation through innovation, underpinned by respect for Dignity, Diversity & Inclusion, and Sustainability.

The ten guiding principles that are set out in the Guidelines broadly follow global trends with a focus on areas such as transparency, fairness and safety. They are divided into two categories with seven principles in the first category applying to all businesses and a further three in the second category expected to be implemented in collaboration with the broader society.

Category One Principles

• Human Centric
• Safety
• Fairness,
• Privacy Protection
• Security
• Transparency
• Accountability.

Category Two Principles

• Education and Literacy
• Fair Competition
• Innovation.

The Guidelines then detail specific measures for each of AI Developers, AI Providers, AI Users, and Advanced AI Systems, with the appendix providing detailed examples and suggestions for each principle and measure accompanied by checklists and sample work sheets.

Public Comments

The draft Guidelines were subject to public consultation, receiving a generally positive response from the industry. However, concerns were raised regarding the document's language, with calls for clarification that the Guidelines set out expectations but do not impose obligations. Intellectual property (IP) issues also attracted criticism, which are understood to be under separate discussion by Intellectual Property Rights Study Group, a working group established by the Intellectual Property Strategy Promotion Secretariat of the Cabinet Office.

Binding Effect and Takeaways

As non-binding recommendations, the Guidelines do not impose penalties for non-compliance. Nevertheless, it could be argued that directors of Japanese companies should adhere to the Guidelines as part of their fiduciary duties. We advise AI operators to follow the Guidelines to the extent that it is commercially reasonable to do so.

Given the broad definition of AI within the Guidelines, a wide array of AI systems may be affected. We recommend that companies seek specific advice on whether their AI programs fall under these definitions and to what extent adherence to the Guidelines is necessary to fulfil directors' fiduciary duties.

 The principles and action items are reasonable and insightful, addressing the key issues which existing/upcoming regulations in the other jurisdictions and thus this is a good reference point to improve corporate governance or ensure compliance with existing laws for other regulated areas such as data privacy.

Next Steps

Given the 'living document' nature of the Guidelines, it is crucial to stay informed about future developments. As regulations and guidance frameworks are being implemented in many countries around the world, organisations will need to take a global and holistic approach to compliance – in particular, organisations will need to pay close attention to regulations which have extra-territorial effect - for example, the EU AI Act, which, whilst being an EU regulation could impact organisations outside of the EU in certain circumstances. (See our additional resources on AI). Additionally, the Nikkei newspaper has reported that the government is considering establishing a binding regulatory framework for major AI developers likely applying only to substantial developers. An approach  focused on regulating key players would be similar to that taken by Japan's Platform Transparency Act  (see our article: Japan's digital platform regulations).

For further assistance or to discuss the implications of the AI Operator Guidelines for your business, please do not hesitate to contact us.