Japan's New AI Law Has Lessons for the US and EU

On June 4, 2025, a Japanese law on artificial intelligence was promulgated, and its title already clearly indicates the political choice made by Tokyo: Jink? chin? kanren gijutsu no kenky? kaihatsu oyobi katsuy? no suishin ni kansuru h?ritsu (Law on the Promotion of Research, Development and Use of Technologies Related to Artificial Intelligence). Unlike the European Union, blocked by the application of a precautionary principle not anchored to objective and measurable elements, Japan has made an extremely pragmatic and conscious choice: not to "regulate AI" — whatever that is — but to enhance what is needed to build the technologies necessary for its functioning.

In short, where the EU applies the handbrake on a parked car with the engine off to prevent it from rolling over, Japan makes sure to build efficient roads to get to its destination faster and better.

“In the past,” explains the Keiyaku Watch observatory , “the EU has promoted the adoption of a so-called hard law, the AI ​​Act, establishing strict rules on those types of AI considered high-risk. In response, the US, fearful that such a choice could slow innovation … coordinated with Japan and other countries to adopt a regulatory approach based on soft law.” However, the EU approach and the change to a “legalistic” course imposed by the Biden administration’s AI Executive Order have convinced Japan to continue along the path of soft law.”

The focus is on applied research and international competition

Article 3 deals precisely with laying the foundations to guide the development of AI in the awareness that competition in this sector knows no borders. Therefore, the rule was born with the aim of improving competitiveness in the international arena of industrial sectors — and it is worth emphasizing the word “industrial” — connected to AI.

At the same time, a brief aside, seemingly placed there by chance, establishes the importance for national security of research and development in AI-related sectors. The importance of this aside lies in the fact that, in a very unhypocritical way compared to the Western debate, it recognizes without dissimulation that AI-related technologies can and must be used for the defense of the country (it is worth remembering that Japan, due to the pacifist nature of the Constitution, cannot have an army with offensive capabilities).

Technological transparency as a tool to prevent illicit activities

Even the issue of illicit uses or uses that damage the regular course of life of the nation is managed in a structural way and not with a myriad of articles that regulate individual cases, with the risk of finding ourselves faced with unforeseen events that cannot be managed in the absence of a specific rule.

The political choice embodied in the law was to focus first and foremost on transparency in every phase of the research, development and deployment cycle of all technologies connected to AI.

This choice deserves further investigation because, unlike the EU regulation on artificial intelligence, it does not impose obligations of “explainability” that are impossible to achieve, but creates the conditions for those who must, and have developed the necessary skills, to be able to verify what has been done, how and by whom.

In other words: mandating by law the “explainability” of AI would imply defining its level. What should be the reference point for measuring explainability? That of a researcher working in a Big Tech? That of a graduate in mathematics? Or that of an ordinary citizen, with a high school diploma?

On the contrary, the obligation of transparency means, much more pragmatically, making it possible for qualified subjects to access all the information necessary to understand what caused damage to private individuals or attacks on institutions.

The scheme of public, private and individual duties

The so-called “architectural” approach of the law on the promotion of AI-related technologies divides duties and responsibilities into three areas.

Unlike EU legislation, Japanese legislation requires all parties to collaborate to achieve the stated goal, i.e., achieving technological leadership. Therefore, central and local public administrations will have to use AI to improve their efficiency, universities will have to actively promote research and dissemination of the results obtained, as well as build a broad and robust knowledge base, cooperating with the State and administrations. Likewise, the private sector will have to improve process efficiency and create new industries through the use of AI-related technologies; and citizens will have to cultivate interest in these technologies.

At the top, it is up to the State to take the necessary measures to ensure that all the actors move in a coordinated manner on this technological stage without hindering each other and without hindering individual performances.

The role of the Prime Minister in implementing the strategy

The strategic approach of this law results in the attribution to the Naikaku— the Prime Minister's Office—of the powers/duties of coordination and control of the way in which the regulatory objectives are implemented. This is done through the establishment of what, in Italy, could be equated to a department of the Presidency of the Council, to which all other State bodies, including independent administrations and agencies, must provide opinions, clarifications and the necessary cooperation.

Access to technological infrastructure and datasets

An extremely interesting aspect of the Japanese AI law is the provision of the obligation to share facilities and equipment — read supercomputing centers, telecommunications networks and more — but above all of datasets to be made available also to the private sector.

This, while in the West — perhaps with the exception of Italy, which in the AI ​​bill proposes a compromise in the name of public interest — we have not yet been able to find a solution to balance, on the one hand, the interests of copyright holders and the (anti-historical) control claims of national data protection authorities and, on the other, the need for access to the resources needed to build models for machine learning and AI.

The geopolitical role of knowledge and the importance of training

The differences between Japanese law and the approach followed by the European Union are also evident with regard to the development of knowledge and training.

Japan clearly recognizes the importance of developing a national knowledge base — that is, not dependent on patents and foreign intellectual property — and, consequently, the need to develop training on AI-related technologies at all levels. Therefore, not only in the field of scientific research but also in those where the results will have to be used.

The importance of strategic vision

No plan can withstand the impact of battle, says a much-used aphorism by General von Moltke, but this does not mean that planning is wrong or impossible. This is precisely the approach that emerges from the setting of the Japanese law on AI, based on the awareness that it makes no sense to cage technological evolution in rules, but that, instead, it is necessary to create an ecosystem that allows its development to be directed by adopting any corrections on a case-by-case basis where they are necessary.

It matters little whether this approach is the fruit of a "political vision" or the consequence of the need to make up for the deficiencies resulting from the decline and aging of the population through automation. In fact, it represents a third way compared to the US one based on "better to apologize than to be allowed" and to that of the EU, which is incapable of freeing itself from bureaucratic dirigism , despite repeated attempts .

It is too early to say which of these approaches will be successful, although the (negative) effects of the first two are already evident. In the US, Big Tech is clamoring for the loosening of restrictions on access to data relating to copyrighted works (subject to legal disputes such as that promoted by the New York Times for the illicit exploitation of its articles) and has launched a massive campaign to obtain the consent (or non-dissent) of users to the reuse of their data .

The European Union is producing cumbersome implementing acts of the AI ​​regulation, nullifying a virtuous approach that uses the rules to support research, which is moreover — in practice — essentially incentivised with economic leverage .

In this regard, if it is true that control over AI technologies is a fundamental element for the development of individual Member States and the acquisition of political autonomy for the European Union, then a strong criticism of its choices is necessary and dutiful. Not because they are necessarily wrong, but because they are read with the glasses of pragmatism and not with those of statements of principle not supported by comparison with reality, but with the claim to bend it.