Part of the AI Regulation News hub.
Japan's Ministry of Internal Affairs and Communications has published the Reiwa 8 Information and Communications White Paper with a special feature on the multifaceted impacts of AI
A white paper binds nobody. This one contains the numbers a Japanese board will be asked about, and one of them says adoption doubled while organisational change did not.
Bottom line: Not binding. This is an annual report to the public on the state of information and communications, the 54th edition since 1973. It contains no rules, no deadlines and no obligations. Its value is evidential.
Who this affects: Executives and strategy teams at companies operating in Japan, and anyone benchmarking Japanese AI adoption against the United States, Germany or China. Policy teams will care because MIC cites the same survey when arguing for policy.
Issue date: 24 July 2026. The full text went up as PDF that day; HTML and EPUB versions are planned for around the end of August 2026, a printed edition from early August, and an English translation later.
What changed: The special feature is devoted to AI, titled on the multifaceted impacts brought about by the advance of AI and oriented toward a digital society in which humans and AI collaborate soundly. It carries a four-country survey of individuals and companies conducted in January and February 2026.
Analysis: The interesting gap is not the adoption gap, which is closing. Japanese companies using generative AI in at least one task reached 86.4 percent against 90.9 in the United States and 91.6 in Germany. The gap that stayed open is organisational: 27.0 percent of Japanese respondents reported no organised effort at all toward transforming work around AI, the highest of the four countries.
Primary sources: MIC press release (JA) · Overview of the 2026 White Paper, annex 1 (JA, PDF) · White Paper landing page (JA)
- Instrument (EN)
- Reiwa 8 Report on the Present State of Information and Communications, known as the 2026 Information and Communications White Paper (working translation)
- Authority
- Ministry of Internal Affairs and Communications, Information Distribution Administration Bureau, ICT Policy Division, Information and Communications Economics Office
- Jurisdiction
- Japan
- Status
- Published; 54th edition since the first in 1973
- Bindingness
- None. An annual public report, not an instrument
- Issue date / next deadline
- 24 July 2026; HTML and EPUB around end of August 2026; English version later, date not stated
- Special feature
- The multifaceted impacts brought about by the advance of AI: toward a digital society in which humans and AI collaborate soundly. Three chapters, on individuals, on companies, and on issues and responses
- Reuse terms
- Free secondary use including reproduction, adaptation and redistribution, except figures on the published exclusion list and text attributed to third parties
- Primary source
- https://www.soumu.go.jp/menu_news/s-news/01tsushin02_02000184.html
The headline number, and the asterisk attached to it
Generative AI use among individuals in Japan reached 58.8 percent in the 2025 fiscal year survey, against 26.7 percent the year before. That is a doubling in twelve months and it is the figure that will be quoted everywhere.
MIC attaches its own caveat, and it matters. The 2023 and 2024 surveys covered people aged 20 and over; the 2025 survey added 15 to 19 year olds, a group whose use rate came in at 91.7 percent in Japan. Restricted to respondents aged 20 and over, the Japanese figure is 52.2 percent.
Both numbers are true. Only one of them is comparable with last year's, and it is not the one in the headline.
Japan is still last of four, and the shape of the gap is specific
Any use of a generative AI service came in at 58.8 percent in Japan, 75.6 percent in both the United States and Germany, and 93.6 percent in China. Broken out by modality, Japanese use of text generation was 55.0 percent, which is not far off the American 65.4 and German 66.0.
Everything else diverges sharply. Image and video generation: Japan 16.3 percent, United States 41.0, Germany 47.6, China 55.8. Voice generation: Japan 6.5 against 29.3 in the United States. Code generation: Japan 7.0 against 35.4.
So the Japanese lag is not a general reluctance. It is concentrated almost entirely outside text.
What people think the thing is
Respondents were asked what kind of presence generative AI is. In Japan, the United States and Germany the most common answer was machine or tool. In China it was secretary or assistant, at 44.7 percent.
The second answers separate. Japan's was dictionary or encyclopedia at 28.9 percent. The American second answer was friend at 25.3 percent, and the German second answer was counsellor at 41.0 percent.
MIC reads this as evidence that in the three higher-adoption countries a substantial group treats generative AI as something to share feelings with, not only as a tool. For anyone drafting an acceptable use policy, that is a more useful finding than any adoption percentage.
The corporate numbers, and where Japan actually falls behind
Companies with a policy of active use or of use limited to defined areas reached 68.9 percent in Japan, up from 49.7 percent a year earlier. Use of generative AI in at least one business task reached 86.4 percent in Japan, and on the same chart 90.9 in the United States, 91.6 in Germany and 98.1 in China. The prior year's Japanese figure was 55.2 percent, so the adoption gap has mostly closed.
Task mix differs. In Japan the top use was assistance with minutes and email drafting, at roughly 70 percent, and it was also where firms most often reported feeling an effect. In the United States use is high across the board and the strongest reported effect was in research and product development.
Now the uncomfortable part. Asked about organised efforts to transform work using generative AI, 27.0 percent of Japanese respondents chose no organised effort, against 1.4 percent in the United States, 4.9 percent in Germany and 2.6 percent in China. On enabling conditions, Japan scored conspicuously low on having an environment to learn the skills and know-how needed, and on training generative AI on internal data or building databases the AI can reference.
Adoption without retrieval infrastructure and without skills provision is adoption of a chat box. The white paper does not put it that way, but the two low bars are the ones a board should ask about.
Two case studies MIC chose to print
Chugai Pharmaceutical appears as the large-company example, with a staged strategy labelled AI Everyday, AI Everywhere and AI Transformation, a digital talent programme, a company-wide idea programme, and an AI governance approach built on human-in-the-loop, meaning human judgment and confirmation are interposed appropriately.
Ainiko Group, a smaller firm in real estate and care services, appears as the counterweight: a one-year focused project starting from each division's own problem statements, reported as saving 2,247 hours of working time annually across ten divisions. MIC notes the firm already had a habit of documenting and improving workflows before any AI arrived, and treats that as the reason the tools landed.
Professor Ichiro Sato of the National Institute of Informatics supplies the framing MIC evidently wanted: management should treat AI as an asset rather than a cost, and use it to turn know-how into an intangible asset rather than to pursue automation as an end.
What we did not verify
The cross-country corporate figures, and the prior-year individual figure, are read off chart labels in the overview annex rather than from its prose. The text layer does not carry them. Anyone quoting 90.9 for the United States should read it from the chart image: the same number appears elsewhere in the extracted text as the complement of a different, individual-use figure, which is a live confusion risk.
I opened the MIC press release of 24 July 2026 and the first twelve pages of the 19-page overview PDF, annex 1. Every percentage, survey detail, case study and expert comment above is taken from those two documents.
I did not open the full white paper, which runs to two parts and a data collection, nor the reuse annex, nor the exclusion list of figures, nor the underlying data files. No English version existed at the time of writing; the release says one will be produced later.
I will not claim any finding from Part II on the state of the ICT sector or on MIC's own policy programme, and I will not treat the survey as nationally representative. It is an online panel survey run through a research company in January and February 2026, with 1,236 individual respondents in Japan and 624 in each of the other three countries, and 515 corporate respondents in Japan and 309 in each of the others. MIC itself notes that attributes such as urban or rural residence were constrained by panel composition, and that people under 15 and aged 70 or over were excluded.
Use this as a benchmark document and cite it carefully. The 58.8 percent headline is not comparable with the prior year because the age base changed; 52.2 percent is. The finding worth acting on is that Japanese firms have closed the adoption gap while leaving the enablement gap open, with 27.0 percent reporting no organised effort at work transformation and notably weak scores on internal data preparation and skills provision. Nothing here binds anyone, which is exactly why it will end up in policy arguments.
Source File
https://www.soumu.go.jp/menu_news/s-news/01tsushin02_02000184.html
Open the overview PDF, annex 1, and confirm three things: the individual use rate chart shows 58.8 percent for the 2025 survey against 26.7 percent for 2024, with a footnote stating the 20-and-over rate was 52.2 percent; the corporate chart shows 86.4 percent of Japanese firms using generative AI in at least one task against 98.1 percent in China; and the work-transformation chart shows 27.0 percent of Japanese respondents selecting no organised effort.
The use rate for generative AI in Japan reached 58.8 percent in the 2025 survey, a substantial rise compared with the 2024 survey figure of 26.7 percent, although it remains lower than in the other three countries. 2026 Information and Communications White Paper overview, Ministry of Internal Affairs and Communications, 24 July 2026, author's translation from the Japanese
FAQ
Is the 58.8 percent figure comparable with last year's 26.7 percent?
Not directly. MIC states that the 2023 and 2024 surveys covered people aged 20 and over, while the 2025 survey added 15 to 19 year olds. On the 20-and-over base the 2025 figure is 52.2 percent.
How far behind other countries is Japan?
On any use of a generative AI service, Japan was at 58.8 percent, the United States and Germany both at 75.6 percent, and China at 93.6 percent. The gap is small for text generation and large for image, voice and code generation.
Does the white paper impose anything on companies?
No. It is an annual public report, the 54th since 1973, and contains no rules or deadlines. It is evidence, not an instrument.
Is there an English version, and can I reuse the charts?
An English translation is planned but was not available at publication. MIC permits free secondary use of the white paper, including reproduction and adaptation, except for figures on its published exclusion list and text attributed to third parties.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.