ChinaCopyright Law, Article 24 (closed list of exceptions, no text-and-data-mining item)
Copyright Law of the People's Republic of China (2020 revision), Art. 24
In force since .
A copyright and text and data mining (TDM) rule binding public and private bodies.
- Audit expectation
- none
What it requires
- Do not rely on Article 24's closed list of exceptions to justify training an AI model on scraped copyrighted content; none of its items covers text-and-data mining or AI training.
- Treat use of copyrighted material beyond personal study, brief quotation, or the other listed exceptions as needing the rightholder's permission.
These lines are LexLint's reading of the law's text.
In the frameworks
- NIST AI RMF
- MAP 4.1Approaches for mapping AI technology and legal risks of its components – including the use of third-party data or software – are in place, followed, and documented, as are risks of infringement of a third party’s intellectual property or other rights.
Our reading of where this law's duties sit in each framework. It never says that running a control meets the law. How we read a framework against the law
Who enforces it
Settledness
- As of
- Open questions
- Is copying works to train an AI model 为个人学习、研究或者欣赏 ("for personal study, research or appreciation") under item 1 of Article 24?
- Does copying works for training 影响该作品的正常使用 ("affect the normal use of the work") or 不合理地损害著作权人的合法权益 ("unreasonably prejudice the lawful rights and interests of the copyright holder")?
What it makes you log
Logging duty
Article 24 lists the uses permitted without the owner's permission and conditions them on naming the author and the work and on not impairing normal use. Naming the author and title is an attribution condition on the use, not a log or record the user keeps. The article imposes no duty to create, keep, produce or review a log or record, and carries no territorial clause of its own, so the hook is empty.
- Kind
- None
- As of
What this law does
Article 24 sets out a closed list of permitted uses that do not require the copyright holder's consent or payment: personal study, brief quotation for commentary, unavoidable reproduction in news reporting, and reprinting current-affairs commentary absent a reservation, among others. There is no general fair-use clause and no text-and-data-mining exception, so training an AI model on scraped copyrighted content has no dedicated statutory basis under this article.
Courts have so far worked the fair-use-adjacent question through AI output cases rather than collection: the Guangzhou Internet Court found AI-generated images infringing in February 2024 (secondary-sourced, the Ultraman case), while the Hangzhou Internet Court held in September 2024, upheld that December, that training itself can be lawful absent intent to plagiarize expression or unreasonable prejudice to the rightholder (secondary-sourced); neither directly answers whether scraping for training is itself permitted.
When LexLint raises it
When your app profile says your app trains models.
Back to the example · Lint your app
Page updated .