Japan Passes Landmark AI Regulation Bill With Risk-Based Framework

Japan Takes a Measured Approach to AI Rules

Japan's National Diet has passed the Act on Artificial Intelligence Governance, making Japan the first major Asian economy to enact comprehensive AI legislation. The law, which takes effect in April 2026, uses a risk-based classification system similar to the EU AI Act but with notable differences designed to preserve Japan's competitiveness in AI development.

Digital Minister Taro Kono said the legislation "balances necessary safeguards with the innovation space our companies need." The bill passed with support from both ruling and opposition parties after eight months of deliberation.

Risk Classification

The law establishes four risk tiers. Unacceptable-risk systems — including real-time biometric surveillance in public spaces and social credit scoring — are banned outright. High-risk systems used in healthcare, criminal justice, critical infrastructure, and employment decisions must meet transparency, explainability, and human oversight requirements.

Medium-risk systems, including customer-facing chatbots and recommendation algorithms, must disclose their AI nature to users. Low-risk systems face no additional requirements beyond existing laws.

Key Differences From EU Approach

Unlike the EU AI Act, Japan's law does not impose mandatory conformity assessments for foundation models. Instead, foundation model developers are required to publish model cards with capability and limitation disclosures and cooperate with a newly created AI Safety Institute for voluntary testing.

This lighter touch on foundation models was lobbied for by Japanese tech companies including NEC, Fujitsu, and Preferred Networks, which argued that overly prescriptive requirements would disadvantage Japanese AI developers competing against US and Chinese companies operating under less restrictive regimes.

Industry Response

Reactions from the AI industry were generally positive. Preferred Networks CEO Toru Nishikawa said the law "provides the clarity we needed without the compliance burden that could have stifled our research." NEC's AI division head Takeshi Yamada welcomed the risk-based approach as "pragmatic and proportionate."

International observers are watching Japan's implementation closely. South Korea and Singapore are both developing AI governance frameworks, and Japan's approach could serve as a model for other Asian economies seeking to regulate AI without constraining their growing tech sectors.