What AI can do, the law can't keep up with. What AI can't do, the law is finally drawing the line.

On September 7, the Supreme People's Court issued the "Opinion on Several Issues Concerning the Application of Law in Adjudicating Artificial Intelligence Dispute Cases." 5 parts, 24 articles, covering AI face-swap infringement, price discrimination, autonomous driving liability, AI-generated content copyright, and more — the hottest and most controversial AI application scenarios. This is China's first AI adjudication rules.

One-sentence summary: when AI causes harm, there's finally a legal framework.

Three Most-Watched Scenarios

AI face-swapping. Over the past two years, the misuse of AI face-swapping technology is no longer news. Deepfake videos used for fraud, pornography, defamation — victims face enormous difficulty seeking justice. The opinion clarifies: using AI to replace someone's facial information without consent constitutes infringement; service providers have audit obligations; platforms bear joint liability. With this line drawn, the legal risk for "AI undressing" and "AI face-swap" apps increases dramatically.

Price discrimination. You and your friend open the same app to book a hotel at the same time, but get different prices — because you searched a few times before, and the system judged you as "price insensitive." This kind of differentiated pricing based on user profiles is explicitly characterized in the opinion as potentially constituting fraud or unfair competition. The impact on platform economics is profound: algorithmic discrimination is no longer a gray area — it's conduct with clear legal consequences.

Autonomous driving. When an autonomous vehicle causes a traffic accident, is the liability on the car owner, the automaker, or the algorithm? The opinion provides a basic framework: for L2 and below (driver assistance), the driver bears primary responsibility; for L3 and above (autonomous driving), the automaker and algorithm provider bear greater responsibility. This grading logic is largely consistent with mainstream international autonomous driving liability frameworks.

AI-Generated Content Copyright: More Complex Than You Think

The opinion also provides guidance on AI-generated content copyright, but this issue is more complex than the previous scenarios.

The core dispute: who owns the copyright for AI-generated text, images, and code? The person who used the AI? The AI developer? Or does it not constitute a work at all, unprotected by copyright?

The opinion's position: it depends on the degree of human participation in the creative process. If humans make substantive creative contributions to AI outputs through prompt design, parameter adjustment, and result curation, then AI-assisted content can be recognized as works, with copyright belonging to the human creator. But if someone simply inputs one sentence and gets generated content, the human contribution isn't "substantive" enough, and it may not constitute a work.

This standard is flexible but brings uncertainty. What counts as "substantive creative contribution"? Does writing a detailed enough prompt count? Does selecting the best image from 100 AI-generated ones count? These questions could spark significant debate in actual trials.

Regulation Has Caught Up With Technology — Now What?

The speed of this opinion's release is fast by judicial document standards. Large AI models have only been in the public eye for two to three years, and the Supreme Court has already produced systematic adjudication rules. This shows how seriously the judicial system takes AI issues.

But the law always lags technology. Today's 24 articles cover today's AI application scenarios. Tomorrow's new technologies, new scenarios, and new disputes will need new rules. The opinion itself leaves room — it references the "technology neutrality" principle, emphasizing that technology development itself cannot be blanketly rejected or blanketly approved.

For businesses, the most practical impact is rising compliance costs. Face-swap apps need auditing, recommendation algorithms need to prove they're not discriminating, autonomous driving needs clear liability allocation — all real financial investments. But long-term, clear rules are good for the industry. When all players know where the red lines are, competition can return to technology and products themselves.

See you tomorrow.