Japan · US — an assembled comparison from each country's own governance data.
Japan and US illustrate different points on the AI governance spectrum. Japan's approach — Soft-law guidelines; active participant in the G7 Hiroshima AI Process; US's approach — Sectoral and executive-led, with significant state-level activity.
On trajectory: Japan is holding a steady regulatory posture; US is loosening its regulatory grip. Most recent tracked marker for each: Japan — "¥1 trillion (~$6.3B) five-year public-private support package announced to build a new domestic foundation-model company, from FY2026." (2026); US — "Newsom signed a package of child-online-safety bills, several aimed directly at AI." (2026-09-10).
Soft-law guidelines; no penalties under the AI Promotion Act
Sectoral, executive-led — no binding federal frontier-model law
Japan (stable): Japan's government policy has generally centered on promotion over restriction, but under PM Takaichi that promotion has become far more concrete — a ¥1 trillion domestic foundation-model push and a government AI platform pilot scaling toward roughly 180,000 civil servants — even as the underlying AI Promotion Act still carries no penalties for non-compliance. Its clearest legislative marker on file is "AI Basic Plan" (in force). US (loosening): The federal posture remains deregulatory on consumer-facing and precautionary rules, but the direction is increasingly asymmetric: on national-security grounds, EO 14409 ('Gold Eagle') and the July agentic-hacking disclosures have pushed the government to assert markedly more control over frontier-model access and cyber-capability testing. With the AI & Crypto Czar role vacant since March, CAISI's standards and testing work remains the most consistent thread of federal activity. Its clearest legislative marker on file is "California SB 53 — Transparency in Frontier Artificial Intelligence Act (TFAIA)" (in force).
¥1 trillion domestic foundation-model push plus the ~$6.3B 'Noetra' robotics/AI consortium
Infrastructure- and export-led; the 'Stargate' initiative and America's AI Action Plan prioritise compute build-out
Japan: High on investment, low on binding rules — state-directed funding, no penalties (notable named developers: Sakana AI, Preferred Networks). US: Low-to-moderate — light federal touch, no binding frontier-model law (notable named developers: OpenAI, Anthropic; AWS leads domestic cloud share at ~28%).
G7 Hiroshima AI Process; UN Independent International Scientific Panel on AI
Chip export-control coordination with allies; no formal multilateral AI safety treaty
No shared "security"-tagged debate axis is recorded for both — on the evidence here, each is legislating its international posture on its own track rather than a shared one.
Among current government priorities: closing the AI investment gap through state-backed industrial strategy
Frontier-model competitiveness and compute infrastructure appear to be higher near-term federal priorities than precautionary regulation, though states continue to pursue their own rules
Where each government is actually spending its attention right now — not a statement of law, but of priority.
4 tracked measures (3 in force, 1 passed)
7 tracked measures (4 in force, 2 passed, 1 proposed)
Japan tracks 5 government AI institutions; US tracks 3 government AI institutions. Raw counts aren't a maturity score by themselves — a single binding statute can matter more than five proposals — but they mark how much formal apparatus each country has stood up so far.
Frontier foundation-model developers, Semiconductor and compute infrastructure
Federal AI safety statute
The gaps recorded for each country don't overlap — each is missing something different, which is a weaker signal than a shared gap but still worth reading as "not present as of the most recent research," not "definitively absent forever."
Shared
Japan
US
Japan and US each describe their own state's role differently: Japan — "High on investment, low on binding rules — state-directed funding, no penalties"; US — "Low-to-moderate — light federal touch, no binding frontier-model law". Layered onto trajectory (Japan stable, US loosening), the open question this comparison actually poses is how much of each country's AI trajectory is being steered by the state versus left to market and research actors — and, for whichever side is tightening, whether that steering is increasing or easing right now.