On July 23, 2025, the Trump administration released its new AI policy, titled Action Plan (2025). While the document includes measures to support businesses, it also touches on data protection—prompting mixed interpretations. Some see it as regulatory tightening, others as a shift toward deregulation, and still others view it as merely a reaction against Biden-era policies.

This ambiguity is not unique to the United States. As countries like the EU, Japan, and South Korea roll out their own AI strategies, what stands out is not clarity, but rather the lack of it.

Beyond growth or regulation, the real question is where governments focus and what they aim to control. Through text analysis, this article explores the contours of their AI strategies.

The survey found that nearly all of Japan’s tunnels have issues, from minor cracks and leaks to serious structural flaws—a troubling reality for a country with vast mountainous terrain and one of the world’s largest tunnel networks.

Actually Cautious: Shades of AI Strategy in U.S. Policy Documents

The Trump administration’s Action Plan (2025) is often viewed as a deregulatory move. Yet a closer look reveals frequent references to “rules,” “laws,” and “standards.” A GPT-based analysis identified 103 regulation-related terms and 42 promotion-related ones, suggesting a clear tilt toward governance.

By contrast, Japan’s *AI Strategy 2022* contains **102 promotion terms** and just **29 regulation terms**, with words like “promote” and “support” dominating. The difference is evident in the comparative bubble chart.

Still, a heavier use of regulation-related language does not necessarily mean a policy is restrictive. In many cases, regulatory design serves as a prerequisite for implementation.

What matters is balance. Japan’s strategy may be ambitious in tone, but the underlying framework appears less developed.

The U.S. Balances Risk and Innovation, Japan Stresses Ethics and Public Value

Looking more closely at the vocabulary, it becomes clearer what each country is actually targeting with its AI policy.

In the U.S. plan, regulation-related terms include “AI risk management,” “accountability,” “safety assessments,” and “regulatory sandbox.” These terms point to **system-level risk control** (shown in blue in the visualization). On the promotion side, terms like “AI innovation,” “open-source AI,” and “AI-driven economy” reflect a focus on **technological leadership and economic development** (shown in red). The dual emphasis on risk and growth suggests a layered strategy, with detailed legal and technical framing.

Japan’s AI Strategy 2022, in contrast, focuses more narrowly on ethics and rights protection, using terms like “cybersecurity” and “privacy” under regulation. Promotion terms reference education, healthcare, and social problem-solving, reflecting a broader public-service orientation and a tone of civic responsibility.

Both countries address AI’s potential and its risks, but the direction and texture of their language reveal key differences in policy maturity and underlying values. (Black terms in the chart represent unique keywords specific to each document.)

From Vision to Implementation: AI Governance Is Already Underway

We also examined how many official AI-related documents are currently in force or soon to be implemented across countries.

The U.S., EU, and Japan have released numerous **guidelines**, indicating a preference for broad principles. In contrast, **China and South Korea are advancing binding laws**, moving toward full regulatory enforcement (see figure). < The U.S. also uses presidential executive orders as an instrument of AI policy.

The content of these documents varies widely. The U.S. focuses on "technical risks" like deepfakes. The EU emphasizes accountability for AI-enabled products. China goes further, addressing technical requirements and oversight obligations, including the structure and output of AI models. Many of these rules are set to take effect after September 2025.

AI policy has moved beyond ethical principles. Countries are now grappling with how to manage deployment at the product level. The U.S. is outlining its vision for 2025. China is methodically building toward 2030. **The global race to operationalize AI is already well underway.

What Japan must catch up with is not vision—but execution. The depth and urgency embedded in policy language say more than ambition alone.