There is now a global consensus on the core principles of intellectual property protection:
Using the intellectual contribution of human beings as the core of rights, building on the training of data compliance as the bottom line, with “who uses and who takes responsibility” as the bounds of authority, while encouraging technological innovation, strictly protects the legitimate intellectual property rights of originalists。
I. Core domestic legal and regulatory framework 1
- the copyright act of the people's republic of china (amended 2020, effective june 2021)
It is a fundamental basis for ai's intellectual property protection
The core sets the core criteria for identifying “human intellectual outcomes + originality”, makes it clear that protected works must be the product of human intellectual creation, and sets down the bottom rules for copyright recognition of ai-generated content。
— civil code of the people's republic of china
Explicitly ai acts of deep synthesis, such as face change, ai emulation, which are not authorized to identify a particular natural person, constitute a violation of the right to portraits, voice, etc., while regulating the liability of ai for producing content that infringes the right to honour, privacy, etc。
2. Generating ai-specific regulatory rules
- interim measures for the management of generating artificial intelligence services (introduced in august 2023)
First-ever ai-specific regulatory document, core intellectual property provisions:
Ai service providers must use training data from legitimate sources and must not infringe on intellectual property rights enjoyed by others in accordance with the law
2 establishes the core principles of “who uses, benefits and who bears the responsibility”
Ai service providers are required to establish mechanisms to deal with complaints of abuse and to protect against the risk of intellectual property abuse。
- the artificial intelligence synthetic content identification scheme (practised in september 2025)
The requirement to produce publicly available ai texts, pictures, audio and video must be clearly marked, with clear responsibilities from generation to dissemination of the whole chain, removal/false markings subject to legal liability, and retention of core evidence for copyright defence of ai content。
3. Ai specific rules for patent protection
Patent review guide (enacted on 1 january 2026)
For the first time, a special chapter was created on the theme “advanced intelligence, big data”, with core rules:
1 add a red line for the review of patent ethics, training data sources in non-compliant ai inventions contrary to good public order, and direct rejection of patent applications
(b) clarify the ai creative standard for patents, the non-licensing of pure algorithms, simple scenario replacements and the need for substantial technical improvements
3 detailed full disclosure requirements for ai patent statements to crack the “black box” model problem。
Uniform rules for the administration of justice (guidance cases + typical cases)
Our country has explored, in judicial practice, the progressive harmonization of ai standards for intellectual property adjudication, with more than 2023-2026 classic jurisprudence defining core rules:
- ai, which produces content that reflects user-specific intellectual inputs (precision of hint design, multi-wheel calibration, content screening modifications, secondary creations), is original and can be protected by copyright law
- enter only the pure ai content of a generic keyword, without human intervention, which does not constitute a work within the meaning of copyright law and is not protected
— protected ai generation content, the copyright of which belongs to the natural/legal person making the substantial intellectual contribution
- unauthorized use of copyrighted works of others for training in ai models constitutes copyright infringement。
Ii. Global core national/regional legislation
1. Eu:
At the heart of the artificial intelligence act (ai act, effective august 2024) and the digital single market copyright directive, the most stringent ai regulatory system around the world has been established。
Core requirements: ai manufacturers must fulfil their training data copyright transparency obligations by making publicly available copyright information for training language materials; ai generation content must be mandatory marking; it is clear that copyrighters can “opt out” their own work for use in ai training data mining。
2. United states:
With case law at its core and adherence to the “human author principle”, the u. S. Copyright agency is unable to register copyrights for clearly pure ai production and can only be protected by copyright if humans make substantial creative contributions to ai production
At the same time, it has been made clear through numerous jurisprudence that the unauthorized use of copyrighted works for training purposes constitutes a violation。
3. Germany:
The adoption in february 2026 of the first ai-generated content force marking act of february 2026 required that all publicly published ai content must be marked with significant and irremovable markings and that offenders face high fines, providing a model for eu ai copyright compliance。
