Why Japan chose to promote AI rather than regulate it — and where that strategy hits its limits
When governments around the world set out to govern artificial intelligence, most of them reach for rules. The European Union has adopted a binding, risk-tiered framework; China steers deployment through centralised directives; and the United States relies on executive orders and agency guidance that shift with each administration. In contrast, Japan took a different route. In May 2025, its National Diet passed the Act on Promotion of Research and Development, and Utilization of AI-related Technology — a law whose organising concept is not regulation or oversight, but promotion.
In a review published in the KeAi journal Risk Sciences, two researchers from Tsinghua University and The University of Tokyo asked why Japan made this choice, and what it has achieved so far.
“The instruments a government chooses reveal its assumptions about the nature of the problem,” explains co-author Hiromi M. Yokoyama from The University of Tokyo's Kavli Institute for the Physics and Mathematics of the Universe. “The urgency of Japan's governance problem comes not from the risks of AI being used, but from the economic and social cost of AI going unused.”
The researchers described a society in which AI adoption lagged well behind other advanced economies, public trust in AI was notably low, and an ageing population coincided with ageing industrial IT systems whose maintenance engineers were approaching retirement. Labour shortages run across agriculture, nursing, construction, retail and logistics. Against this background, AI is cast as a technology that can compensate for labour the demographic structure no longer supplies.
One finding stood out. Cross-national surveys by the team show that Japanese respondents score among the highest on objective tests of AI knowledge, yet rank lowest in confidence about their own understanding. Japanese reluctance, the authors note, concentrates in trust rather than acceptance.
The study also explains why promotion could work at all in Japan. The country's tradition of administrative guidance – in which agencies steer private behaviour through non-binding recommendation – together with a corporate culture attentive to compliance and reputation, and close coordination between ministries and organised business, allows voluntary guidelines to function as real constraints without legal compulsion. The Act's central organisational instrument, an AI Strategy Headquarters chaired by the Prime Minister, is designed to overcome the fragmentation of authority across ministries that has hampered Japanese policy in other fields.
As to whether it has worked, in the period since the Act was passed, personal use of generative AI rose sharply and trial use within firms approached international levels. However, the authors identified a second dimension that promotional instruments do not reach. That is, the gap between large firms and small and medium enterprises in actual deployment has not narrowed, and self-assessed transformation maturity remains near the bottom of the scale.
"Adoption is not a single quantity," says co-author Shuo Wang of Tsinghua University's School of Marxism. "Trial and organisational integration respond to different instruments. Promotional instruments lower the cost of trying. They do not supply the capital, technical personnel or managerial authority that system replacement and process redesign require — and these are precisely what smaller firms lack."
The Japanese government appears to have reached a similar conclusion: its second AI Basic Plan, adopted in July 2026, shifts the objective to organisational change in how decisions are made and work is organised. The authors also cautioned that voluntary compliance offers weak protection against rare, high-impact events, and that codifying a promotional orientation in statute may create path dependencies that resist regulatory instruments when circumstances warrant them. Yet the annual revision built into the Basic Plan mechanism — which held through a change of premiership — offers a channel for adjustment that does not require reopening the law.
“A governance model may succeed on its own terms while the underlying problem persists in a form its instruments cannot readily reach.” Yokoyama adds. “Japan's case shows that where adoption is low, the governance problem has a different structure from the innovation-versus-risk balance that dominates comparative debate, and it calls for different instruments.”
###
Contact the author: Hiromi M. Yokoyama, Center for Data-Driven Discovery (CD3), Kavli Institute for the Physics and Mathematics of the Universe (Kavli IPMU, WPI), The University of Tokyo, Kashiwa, Japan. hiromi.yokoyama@ipmu.jp
The publisher KeAi was established by Elsevier and China Science Publishing & Media Ltd to unfold quality research globally. In 2013, our focus shifted to open access publishing. We now proudly publish more than 300 world-class, open access, English language journals, spanning all scientific disciplines. Many of these are titles we publish in partnership with prestigious societies and academic institutions, such as the National Natural Science Foundation of China (NSFC).
Journal
Risk Sciences
Method of Research
Data/statistical analysis
Subject of Research
Not applicable
Article Title
"Promotion" as a policy instrument: The social roots and implications of Japan's light-touch AI governance
