China issues national guidelines for court handling of AI disputes

China’s Supreme People’s Court has released a comprehensive set of guidelines to standardize how courts handle legal disputes involving artificial intelligence. The document, issued on September 7, provides a national framework for assessing liability in cases where AI systems cause harm, such as through deepfakes, privacy breaches, or defamatory outputs. This move aims to resolve inconsistencies that have arisen as Chinese courts have faced a growing number of AI-related cases in recent years.

The guidelines do not create new legislation but instead offer interpretive directions for applying existing laws, including the Civil Code and the Personal Information Protection Law. By establishing a unified approach, the court seeks to bring clarity to a legal landscape that has previously seen varying outcomes across different jurisdictions. The guidance covers a wide range of issues, from personality rights to the use of AI in autonomous vehicles and judicial proceedings.

Victims must prove fault for AI harms

A central feature of the new guidance is the establishment of ordinary fault liability as the default rule for AI-related torts. Under this standard, the party claiming harm must demonstrate that the defendant was at fault. This differs from strict liability regimes, where the defendant is held responsible regardless of intent or care. The court stated that this approach is intended to prevent excessive legal burdens on developers and providers during an early stage of technological development.

For consumers and victims, this creates a significant hurdle. They must now prove that the AI provider failed to take reasonable precautions or that the system’s autonomy and transparency contributed to the harm. The guidelines direct courts to consider specific factors, such as the level of system autonomy, the risks involved, and the measures taken to prevent damage. This trade-off protects industry innovation but places a heavier evidentiary burden on individuals seeking compensation.

Human actors remain legally responsible

The guidelines explicitly reject the idea of AI as a legal actor. The system itself cannot be held liable; instead, responsibility falls on the humans and entities involved in developing, providing, or using the technology. This means that companies and individuals remain accountable for the outputs generated by their AI tools. The fact that a machine produced the harmful content does not serve as a defense for the parties who deployed it.

This stance aligns with a broader global trend of holding developers and users accountable for AI behavior. However, it also highlights a gap in the current legal framework. As AI systems become more autonomous, determining the precise point at which human control ends and machine autonomy begins remains a complex legal challenge. The guidelines provide a starting point but leave room for interpretation in edge cases.

Copyright and training data remain unsettled

Despite the broad scope of the guidelines, key questions regarding intellectual property and model training data remain unresolved. The document does not provide a definitive answer on whether using copyrighted material to train AI models constitutes infringement. This omission reflects the ongoing debate over the legality of training data usage, a topic that continues to generate litigation in various jurisdictions.

According to GN technics/ai (en-US), the guidelines serve as a judicial policy document rather than formal legislation. This legal status allows for flexibility but also means the rules can evolve as technology changes. China has not yet enacted a comprehensive AI law, so these guidelines act as a critical bridge between existing statutes and the rapid advancement of artificial intelligence. The approach balances the need for legal certainty with the imperative to foster technological growth.