China's Supreme Court Sets National Standard for AI Liability

China's Supreme People's Court has issued comprehensive judicial guidance establishing how courts nationwide should handle disputes involving artificial intelligence, marking a significant shift in how the world's second-largest economy approaches AI liability.

The Opinions on Adjudicating Disputes Involving Artificial Intelligence in Accordance with the Law, released September 7, contain 24 provisions covering personality rights, personal information, intellectual property, consumer protection, autonomous vehicles, and AI use in judicial proceedings. The guidance comes as Chinese courts handled 908 cases involving data ownership and transactions in 2025 alone, representing a 25.6% increase from the previous year.

Why it matters

Unlike the EU's comprehensive legislative approach or the U.S.'s case-by-case litigation model, China is establishing a unified judicial framework for AI disputes before enacting comprehensive AI legislation. For multinational companies, this creates a third distinct liability regime to navigate, with practical implications for how they document training data, implement safeguards, and respond to infringement complaints across borders.

Fault liability becomes the default

The court established ordinary fault liability as the general rule for AI-related torts, requiring plaintiffs to prove a defendant's fault rather than placing the burden on defendants to prove they weren't at fault. The Supreme People's Court stated this approach aims to avoid imposing excessive liability that could discourage innovation while AI technology remains in early development stages.

Courts will consider AI-specific factors when determining fault, including system autonomy, technological transparency, potential risks, preventive measures taken, and parties' ability to foresee and control risks. Critically, the guidance makes clear that AI itself bears no legal responsibility—liability rests with the people and entities developing, providing, or using AI systems.

Training data and generated content

For AI model training, processing personal information that individuals have made public generally won't constitute infringement if it remains within reasonable scope and the individual hasn't expressly objected. However, consent is required where processing significantly affects individual rights.

Generative AI providers face potential liability if they receive proper notice that their service generates content infringing personality rights and fail to take prompt action. Users who deliberately craft prompts to induce harmful output may also be liable.

In copyright disputes, developers asserting noninfringement may be required to produce evidence concerning training-data sources, training records, and model operations. This requirement addresses evidentiary challenges created by opaque AI systems, making documentation of training-data provenance and model development important for both compliance and litigation defense.

Notably, the court left two major copyright questions unresolved: whether AI-generated content qualifies for copyright protection and how unauthorized use of copyrighted works for model training should be treated, citing significant disagreement and the need for further judicial experience.

AI in the courtroom

The guidance also addresses generative AI use by lawyers and litigants. Parties using AI to prepare court filings or research reports must verify the authenticity and accuracy of cited materials, disclose their AI use, and remain responsible for submitted content. Using AI to fabricate evidence or create false litigation may result in fines, detention, or criminal liability.

Broader enforcement coordination

The Supreme People's Court called for coordination among courts, cyberspace regulators, public security authorities, procuratorates, and market regulators, signaling that AI governance extends well beyond civil litigation. The guidance emphasizes that AI misconduct may trigger civil remedies, administrative penalties, and criminal liability.

These details were first reported by the International Association of Privacy Professionals.