Is Web Scraping Legal in Japan? A 2026 Guide

Most legal guides to web scraping default to US law — the CFAA, hiQ v. LinkedIn, Terms of Service as contract. Japan's rules are different, and in one respect the difference is stark: Japan's copyright law is more permissive toward scraping for data analysis than the US, while a separate criminal statute makes request rate — not copying — the thing that actually gets people arrested. If you're scraping Japanese sites, or a Japan-based company evaluating scraping, the US framing will point you at the wrong risks.

The two questions that actually decide it in Japan

Skip the US mental model. In Japan, a scraping project mostly turns on:

Request pacing is the Japan-specific risk that most English-language legal guides never mention. It's covered second, below — it's arguably the more practical one.

Copyright: Article 30-4 is unusually permissive

Japan's Copyright Act Article 30-4 creates a broad exception: you may copy a copyrighted work, by any method, to the extent necessary, when you are not letting yourself or anyone else enjoy ("kyōju") the thoughts or feelings expressed in it. "Jōhō kaiseki" (information analysis) — extracting language, sound, or image data from a large number of works and comparing, classifying, or otherwise analyzing it — is the article's headline example. In practice, this covers most data-analysis and ML-training scraping: you're extracting facts and patterns from pages, not distributing the pages' creative content for people to read.

This is a genuinely different starting point than US fair use, which is a four-factor balancing test decided case by case. Article 30-4 is closer to a categorical rule: if the use is non-enjoyment (analysis, indexing, training), it's presumptively fine.

The exception's limit: the 但書 (proviso)

Article 30-4 has a proviso — it doesn't apply if the use "unreasonably prejudices the interests of the copyright holder," judged by the type and use of the work and how it's used. Japan's Agency for Cultural Affairs (文化庁) published official guidance in March 2024 ("AI and Copyright," 令和6年3月) with three concrete examples of what crosses that line:

1. Idea-level similarity alone is fine. If outputs merely resemble the ideas in training works — not their protected creative expression — mass-generating similar-feeling content is not "unreasonable prejudice," even if it competes with the original creators.
2. Scraping a paid database instead of paying for its API is the clearest violation. The Agency's example: a site sells API access to a curated, analysis-ready database of its articles. Scraping the same articles from the public pages instead of paying for the API — when the scraped data functions as a substitute for the paid database — falls outside the exception.
3. Circumventing anti-scraping measures aimed at a future paid product is also a violation — if a site has technical measures blocking bulk collection, and there's a reasonable inference that a licensable database drawing on that site's content is planned, defeating those measures to collect the data anyway can foreclose that future market.

The practical read: analyzing public web data is Article 30-4's core case. The exception erodes specifically when a paid, structured alternative exists and you route around it, or when you defeat technical protections aimed at a data product the site is building.

The bigger risk: 偽計業務妨害 (obstruction of business)

Copyright isn't where scraping cases in Japan actually get people arrested. Penal Code Article 233 criminalizes "obstruction of business by fraudulent or deceptive means" (偽計業務妨害) — and it has been applied to ordinary crawling.

The lesson isn't "crawling is illegal in Japan" — it's that the request pattern itself, not the copying or the downstream use, is the thing Japanese authorities have treated as potentially criminal, especially against smaller or public-sector infrastructure. A rate that would draw a polite cease-and-desist letter in the US drew a criminal arrest in Japan. Pace conservatively, respect robots.txt and any documented rate limits, and prefer targets built to handle programmatic traffic.

Personal data: the APPI still applies to public pages

Japan's Act on the Protection of Personal Information (個人情報保護法, APPI) treats information as "personal data" whenever it identifies a living individual — being publicly visible on a web page doesn't exempt it. Two practical consequences for scraping:

• 要配慮個人情報 (special-care-required personal data — race, health, criminal history, and similar categories) generally cannot be collected without the person's consent, full stop.
• Publishing or reselling personal data you scraped — passing it to a third party as-collected — is generally a "third-party provision" requiring consent, even if the original source was public.

The safest path is the same one that works everywhere: prefer business- and product-level facts (prices, listings, ratings, company data) over records that identify individuals, and treat any personal-data collection as its own compliance question.

Commercial databases: the Unfair Competition Prevention Act

A 2018 revision to Japan's Unfair Competition Prevention Act (不正競争防止法) created a protected category called "限定提供データ" (limited-provision data): data that a business (1) provides to specific parties through paid transactions, (2) has accumulated in meaningful volume electronically, and (3) access-restricts (typically via ID/password or similar gating). Wrongfully acquiring or using that kind of data — for example, scraping large volumes of it from behind its access controls and reusing it in a competing service — is "unfair competition," and the data owner can seek an injunction.

This sits alongside copyright, not instead of it: a paid, access-gated commercial dataset can be protected under the Unfair Competition Prevention Act even for facts and compilations that copyright wouldn't cover on their own. It's the same instinct as Article 30-4's proviso — routing around a paid, gated product is where the exposure concentrates — approached from a different statute.

A practical checklist for scraping in Japan

Lower risk

• Public, non-personal, factual data (prices, catalog listings, business info).
• Conservative request pacing — well below anything that could degrade the target, especially for smaller or public-sector sites.
• Analysis and internal use of what you collect, not republishing the copyrighted expression itself.
• Free, publicly accessible pages rather than a paid tier's underlying database.

Higher risk

• High-frequency requests against infrastructure that can't obviously absorb them.
• Special-care-required personal data, or reselling/publishing personal data you collected.
• Scraping the public-page equivalent of a site's paid API or access-gated database.
• Defeating technical anti-scraping measures aimed at a data product the site is building or already sells.

How Crawlora fits

Crawlora's platform APIs return normalized JSON for public sources through managed, rate-limited infrastructure — the request-pacing discipline that Article 233 exposure specifically turns on is handled for you, rather than left to a DIY script hammering a target as fast as it can. You're still responsible for what data you collect and what you do with it; see is web scraping legal in 2026 for the equivalent US-law breakdown, and how to check if a site prohibits scraping before you start.

Build on managed, rate-limited infrastructure

Documented endpoints, normalized JSON, and paced access by design. 2,000 free credits a month, no card.

Try the PlaygroundRead the docs

Related reading

• Is Web Scraping Legal in 2026? A Practical Guide — the equivalent US-law breakdown (CFAA, hiQ, GDPR/CCPA).
• Scraping Sites That Block Bots — the technical side of what Article 30-4's proviso calls "technical measures."
• Best Web Scraping APIs in 2026: How to Choose — infrastructure that handles pacing for you.