On May 27, 2026, Liu Guixiang, a standing member of the Supreme People's Court's Adjudication Committee, told a State Council Information Office briefing that Chinese courts would spend the 15th Five-Year Plan period (2026-2030) drafting new judicial guidelines for "data rights, data transactions and AI-generated content" (SPC, court.gov.cn). Read plainly, it is an unusually candid admission: four years into the generative-AI boom, China's courts still have no unified national standard for adjudicating AI disputes — and won't have one for another four.
The Patchwork the SPC Is Trying to Fix
The gap is not theoretical. Since November 2023, when the Beijing Internet Court issued China's first ruling recognizing copyright in an AI-generated image, at least five courts have reached materially different conclusions on similar facts. The Hangzhou Internet Court held a generative-AI platform contributorily liable for outputting Ultraman-lookalike images, a ruling the Hangzhou Intermediate Court upheld on appeal in December 2024. The Changshu People's Court recognized copyright in a Midjourney-assisted image in March 2025. Days apart that same month, the Zhangjiagang Court went the other way, ruling that content "autonomously created by AI" is not a "work" under copyright law at all — the first Chinese decision to explicitly deny protection on those grounds (IAM Media). By September 2025, the Beijing Internet Court had settled on a middle position requiring proof of "personalized expression" and creative effort. Four courts, four fact patterns, three distinguishable standards — and no case has yet reached the Supreme People's Court itself to consolidate them.
Data rights are similarly unsettled. China built a "three-rights" framework for data property in 2022 (holding, processing, and product-management rights), and eight provinces are running data-intellectual-property pilots, but there is still no single national standard for who owns a dataset, a data product, or the output of a model trained on both. Liu's announcement is effectively an acknowledgment that the SPC will now try to knit those local experiments into binding doctrine — a five-year construction project, not a rule that exists today.
The Judge-in-the-Loop Baseline Already Exists
It would be wrong to read this as China starting from zero. The SPC's earlier "Opinions on Regulating and Strengthening the Application of Artificial Intelligence in Judicial Fields," issued December 12, 2022, already sets the guardrail that matters most: AI tools may generate "supplemental references" for a case, but "AI shall not make judicial decision[s] substituting for the judge in any case, disregarding technology advancement" (court.gov.cn, 2022). That opinion targeted courts' use of AI in casework — sentencing aids, document drafting, case-management triage. The 2026 announcement targets something harder: how courts should rule on disputes about AI products built by others. The first problem — keeping a human in the loop — China solved early and clearly. The second — giving litigants and AI firms predictable rules for who owns what and who is liable for what — it has not.
The Case for Patience, and Its Limits
There is a real argument for the SPC's slow-build approach. Common-law systems, including the United States, often let case law mature through appellate review before legislating, on the theory that judges need a body of fact patterns to generalize from rather than guessing at hypotheticals. China's courts are still surfacing genuinely hard line-drawing questions — how much human "selection and embellishment" of a Midjourney output should earn copyright, for instance — that a rushed national rule could get wrong in ways iterated litigation would not. Data property is a harder doctrinal problem still, since it requires reconciling ownership, privacy, and national data-security law at once; getting that architecture wrong at the top could freeze the very data-element market the Five-Year Plan is trying to grow.
But four years of case-by-case drift already has costs a five-year runway will compound. Generative-AI platforms operating nationally cannot price litigation risk when Zhangjiagang and Changshu courts disagree on whether their outputs are copyrightable at all. That uncertainty falls hardest on smaller Chinese AI developers who cannot absorb split-jurisdiction risk the way Alibaba or ByteDance can — precisely the "new quality productive forces" Beijing says it wants to cultivate. Tellingly, the State Council's parallel 2026 legislative work plan is already accelerating a first comprehensive AI law covering data, algorithms, and IP together, which suggests Beijing itself sees the current patchwork as untenable now, not in 2030 (The Legal Wire).
A Narrower, Faster Fix Would Serve Innovation Better
The SPC does not need the full five-year plan to close the gap that matters most to industry. It could issue interim guiding cases — a tool Chinese courts already use to bind lower courts without full legislation — on the narrow copyright-threshold question dividing Zhangjiagang and Beijing, well before 2030. Broader data-property doctrine can reasonably take longer given its entanglement with national security law. As Oxford's Institute for Technology and Justice has noted, courts adjudicating AI matters without any nationwide framework remains the norm across China's justice system, not just in civil IP cases (Oxford BSG). Proportionate regulation does not mean unhurried regulation. Where courts have already generated enough conflicting precedent to identify the actual disputed question, the SPC should resolve it years before its own deadline — not at it.