AI and Copyright: Comparing Legal Frameworks in Japan and the US

About 10 minutes

Legal and IP teams at companies using generative AI, AI product developers, business professionals seeking to understand copyright basics in the AI context

AI Governance Overview

AI and copyright is one of the most pressing legal risks facing organizations that use generative AI. As tools that generate images, text, and code become commonplace in business operations, the legal status of training data and ownership of AI outputs continues to be actively debated in both Japan and the United States.[1][3][4]

This article examines the current state of copyright law in both countries, summarizes recent developments, and outlines practical steps companies can take. Since legal frameworks in this area continue to evolve, I recommend consulting a specialist for decisions specific to your situation.

Basic Copyright Concepts in the AI Context

What Constitutes a Copyrighted Work

Copyright law defines a “copyrighted work” as the creative expression of human thought or emotion. In Japan, this is defined under Article 2, Paragraph 1 of the Copyright Act; in the United States, under 17 U.S.C. § 102.[1][2]

In both countries, a key shared principle is that human creative contribution is required for copyright protection to apply. Whether content generated autonomously by AI qualifies as a copyrightable work is currently under active discussion in both jurisdictions.[3][5]

The Copyright Attribution Question

AI-related copyright issues generally fall into two distinct contexts:

Training phase: Whether using existing copyrighted works (images, text, code, etc.) to train AI models is legally permissible
Generation phase: Whether AI-generated outputs can be protected by copyright, and if so, who holds those rights

These two contexts raise different legal questions.

The Legal Framework in Japan

Current Copyright Act Provisions

Japan’s Copyright Act was amended in 2018 to introduce a specific exception for information analysis purposes.[1]

Article 30-4 (Use of works for purposes other than enjoying the thoughts or emotions expressed therein) permits reproduction and adaptation of copyrighted works without the rightsholder’s authorization, provided the purpose is information analysis, machine learning, or similar — not the “enjoyment” of the work’s expressive content.[1]

An important caveat is that uses that “unreasonably harm the interests of the copyright holder” are excluded. The interpretation of this condition is a key point of practical discussion.

Article 47-5 permits the provision of minor portions of works in the context of information analysis services, and serves as a supplementary basis for AI-powered information processing services.[1]

Using Copyrighted Works as AI Training Data

Under Article 30-4, Japan generally permits the use of copyrighted works as training data for information analysis purposes without prior authorization. On the face of the statute, this represents a broader scope of permissible use than the explicit fair use provisions in the United States.[1][4] That said, several areas require attention:

Restrictions imposed by the terms of service of the data source
Uses that unreasonably harm the copyright holder (e.g., large-scale acquisition and redistribution)
Prohibition on circumventing technical protection measures

Copyright Ownership of AI-Generated Outputs

Under current Japanese copyright law, protection requires “the creative expression of a person.” For this reason, content generated entirely autonomously by AI may not qualify as a copyrighted work.[3]

On the other hand, where a human has made a creative contribution — such as prompt design, selection, or editing — there is room for the output to be treated as a work authored by that person. As of June 2026, however, there are few established precedents defining the threshold for what level of human involvement qualifies, and interpretation continues.[3][7]

Recent Developments

In March 2024, the Agency for Cultural Affairs published “Thinking on AI and Copyright.” This document clarifies how copyright law applies to both the training and generation phases of AI, and serves as a primary practical reference.[3]

The Legal Framework in the United States

The Fair Use Doctrine and AI Training

The U.S. Copyright Act (17 U.S.C. § 107) codifies the fair use doctrine. Fair use determinations weigh the following four factors:[4]

The purpose and character of the use (commercial or transformative)
The nature of the copyrighted work
The amount and substantiality of the portion used
The effect of the use on the potential market for the original work

Whether using copyrighted works as AI training data constitutes fair use is currently being contested in ongoing litigation. As of June 2026, no settled legal standard has been established.[6][8]

Copyright Office Policy

The U.S. Copyright Office requires “human authorship” as a condition for copyright registration.

In the February 2023 “Zarya of the Dawn” case, the Copyright Office declined to register the AI-generated image portions of a work created with Midjourney. The human-authored portions (such as text) were accepted for registration, while the AI-generated portions were excluded from protection.[7]

The Copyright Office published guidance in March 2023 on registering works that contain AI-generated content. Additional policy documents have been issued through 2024 as the Office continues to refine its approach.[5]

Key Litigation Cases

Getty Images v. Stability AI (2023–present) Getty Images filed suit against Stability AI in both the United Kingdom and the United States, alleging that its photographs were used as training data without authorization. This is one of the most closely watched cases addressing the legality of training data.[8]

The New York Times v. OpenAI and Microsoft (filed December 2023) The New York Times sued over the alleged unauthorized use of its articles to train ChatGPT. The complaint includes specific examples demonstrating the model’s ability to reproduce article content verbatim, raising significant questions about the scope of “transformative use.”[6]

Japan–US Comparison Summary

Aspect	Japan	United States
AI training use	Generally permitted under Article 30-4 (with the exception for unreasonable harm)[1]	Determined case-by-case under fair use; major litigation ongoing[4][6][8]
Copyright in AI outputs	Human creative contribution required; autonomous AI output unlikely to be protected under current law	Human authorship required; Copyright Office policy excludes AI-generated portions from registration
Primary guidance	Agency for Cultural Affairs, “Thinking on AI and Copyright” (March 2024)	U.S. Copyright Office guidance on AI and copyright (March 2023–)
Legal certainty	Statutory basis is relatively clear, but case law is limited[1][3]	Fair use requires individual analysis; outcomes of key cases still pending[4][6][8]
Litigation landscape	Significant domestic litigation limited as of June 2026[3]	Multiple major cases in progress; initial rulings have begun[6][8]

Practical Steps for Companies

1. Training Data Governance

Organizations that develop or fine-tune AI models need a process for verifying the legal status of their training data.

Review the provenance of data and applicable terms of service
Prioritize datasets with clear licenses
Maintain records of data sources, acquisition dates, and license terms

Even when using third-party AI services (APIs or SaaS), reviewing that service’s training data policy and understanding where legal risk may reside is a useful practice.

2. Establishing a Policy for AI-Generated Content

Defining how AI-generated content will be handled internally is a practical starting point for risk management.

Scope of permitted use (e.g., external publication, commercial use)
Policy on copyright attribution (whether human creative involvement is documented)
Review of how the terms of service of the AI tool define ownership of outputs

3. Risk Management and Contract Review

Incorporating copyright considerations into relevant contracts is worth considering.

Adding warranty provisions related to AI-generated content in vendor agreements
Disclosing and obtaining consent when deliverables to clients include AI-generated content
Reviewing indemnification clauses related to intellectual property

4. Internal Education and Guidelines

As more employees use AI tools in their daily work, building a baseline understanding of copyright across the organization becomes important.

Creating internal guidelines covering copyright basics and considerations specific to AI use
Establishing a process for involving legal or IP teams when new AI tools are introduced
Setting up a system for staying current with changes in law and regulation

Summary

In both Japan and the United States, the legal landscape around AI and copyright remains in flux. Japan has established a relatively clear statutory basis under Article 30-4 for training-related use, but questions about the protection of AI-generated outputs are still being interpreted. In the United States, cases are decided on an individual fair use analysis, and the outcomes of major ongoing litigation are expected to shape future practice.[1][3][4][6][8]

Steps companies can take now include establishing training data governance, defining policies for AI-generated content, and reviewing contracts and internal education programs. Monitoring changes in the law and addressing these areas incrementally is a realistic approach given the current state of uncertainty.

For decisions that require specific legal judgment, consulting a lawyer or IP attorney who specializes in intellectual property is recommended.

References

e-Gov Law Search, Copyright Act, Articles 2, 30-4, and 47-5
U.S. House of Representatives, 17 U.S.C. § 102 - Subject matter of copyright
U.S. House of Representatives, 17 U.S.C. § 107 - Limitations on exclusive rights: Fair use
U.S. Copyright Office, Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, March 16, 2023
The New York Times Company, Complaint: The New York Times Company v. Microsoft Corporation and OpenAI, December 2023
U.S. Copyright Office, Re: Zarya of the Dawn (Registration # VAu001480196), February 21, 2023
The Verge, Stability AI’s legal win over Getty leaves copyright law in limbo, November 5, 2025

Quiz

Generative AI and Personal Information: Legal Obligations and Practical Compliance

Human-in-the-Loop vs. Human-over-the-Loop