Summary of the Key Points
The recent judicial opinion issued by the Supreme People's Court regarding intellectual property rights in the field of artificial intelligence (AI) essentially provides a practical "transitional set of rules" to address the current chaos in AI-related copyright and patent disputes. On one hand, it clarifies the responsibilities that have already been agreed upon in the industry, addressing urgent issues such as the lack of liable parties in cases of AI infringement, the reluctance of open-source developers to contribute code for free, and the inability to patent new inventions created by AI. On the other hand, it intentionally leaves open two globally controversial issues without making definitive conclusions, avoiding the risk of either stifling the AI industry or compromising the interests of ordinary creators, thus allowing for sufficient room for trial and error in technological development.
---
Detailed Analysis of the Rules
1. AI Infringement: No Longer Blame-Shifting
Previously, when AI infringements occurred, it was often difficult to determine who was responsible. For example, if an AI-generated poster copied a designer's original artwork, the company that developed the AI model might claim it only provided the underlying technology, the platform that sold the AI drawing tool might argue that the user's input was responsible, and the business that used the AI to create the poster might say it used a legitimate AI tool and thus should not be held accountable. This time, a comprehensive accountability system has been established. Courts will consider the type of AI, the source of training data, the contribution each party made, and the profits earned, determining who is at fault.
The most significant change is the mandatory obligation for AI developers to prove their innocence. If someone claims their AI did not infringe, they cannot use the argument that "AI is a black box and they cannot explain how it works" as a defense. They must provide detailed information about the training data, the entire training process, and the logic of the model, effectively exposing the previously hidden "black box" and making it impossible to evade responsibility.
2. AI-Invented Patents
There has been long debate in the industry about whether inventions created with AI are protected by patents. For instance, whether a new drug formula or chip design developed with AI assistance qualifies as a patentable invention. The new rules state that as long as the AI-generated result follows natural laws and solves a technical problem (for example, if a new drug formula is derived with AI assistance in just three months), it can be patented, provided it meets certain criteria. However, there are two key restrictions: the inventor must be a natural person who has made a creative contribution; AI itself cannot be recognized as the inventor. Additionally, the patent application must clearly describe the technical solution so that it can be replicated by anyone in the industry. Using the argument that the AI-generated logic is unclear as a reason to avoid patenting is not acceptable.
3. Open-Source Code: Developers Exempt from Liability
More than 90% of major AI models in China rely on free, open-source code. Developers were concerned that contributing such code could lead to liability in case of infringement. The new rules reassure them that if they release open-source AI code for free, do not profit from it, and clearly document its functionality and potential security risks, they are not responsible for any subsequent infringement. This encouragement has significantly boosted innovation in the open-source community.
4. Two Controversial Issues Reserved for Further Discussion
Two highly debated issues were not resolved in the opinion, out of practical considerations:
- Copyright for AI-Generated Content: Whether content created by AI, such as paintings or novels, is protected by copyright. Currently, there is no global consensus. The rules allow courts to decide on a case-by-case basis based on the specific circumstances.
- Using Others' Content for AI Training: Whether using others' novels, news, or images in AI training constitutes infringement. Proponents argue that it is similar to learning from books and that requiring AI companies to obtain licenses from each author would be unfeasible and hinder AI development. Opponents argue that AI can create content that closely resembles the original work, potentially infringing on authors' rights. These issues will be addressed later as more cases are adjudicated and more experience is accumulated.
5. A Balanced Approach to Regulating the AI Industry
There were concerns that the Supreme People's Court's rules might either favor AI companies at the expense of creators or overly restrict AI innovation. The new rules strike a middle ground, clearly defining what is allowed and what is not, providing a clear safety boundary for all stakeholders. This allows companies, creators, and open-source developers to anticipate the consequences of their actions without constant fear of legal disputes.