On 22 April 2026 the Beijing High People’s Court released its 2025 Typical Cases on Judicial Protection of New Quality Productive Forces, including a case on the crime of copyright infringement using AI generative models, which the court highlighted at the release.
The facts and the ruling
According to public information, the defendants, without a licence, used open-source generative models and other tools to process others’ original artwork and sold the results through e-commerce shops for profit. The first-instance court held the defendant entity and four individual defendants guilty of the crime of copyright infringement, and the judgment is now effective.
The court stressed that AI is not a shield against infringement: using AI to copy and distribute others’ original artwork for profit meets the elements of the crime. The case also clarified the standards for identifying unit crime and joint crime, targeting the full “steal image — AI-process — e-commerce sale” chain.
Why criminal, not civil
The significance of the case is that it pushes AI infringement from the civil level to the criminal level.
On the civil path, the rights holder seeks cessation and damages; on the criminal path, conduct reaching “for-profit, unauthorised copying and distribution” and of a serious degree may constitute the crime of copyright infringement. The thresholds and consequences differ. Choosing criminal filing means the authorities consider the scale and subjective culpability to warrant criminal intervention.
The “full chain” language matters. The court does not look only at the party doing the processing but evaluates stealing the material, AI processing and e-commerce sale as a whole, so participants along the chain may all be pursued. This connects directly to identifying unit crime and joint crime.
Three risk lines for companies
First, source of material. The first step in AI-assisted creation or processing is confirming the lawful source of inputs. Using others’ works on the basis that “it can be found online” is the highest risk. See our IP protection service.
Second, characterisation of processing and output. If the AI-processed product substantially reproduces the original’s original expression, it may infringe even after style transfer or partial modification. The test is substantial similarity, not pixel-level identity.
Third, sale and dissemination. Large-scale, for-profit copying and distribution is the core trigger of criminal risk. Operators of e-commerce, livestreaming and content-distribution channels have a duty to verify the copyright source of goods and material — see our criminal compliance service.
Suggested compliance actions
First, keep a material ledger. Record source, authorisation evidence and scope of use, so it is traceable.
Second, add pre-listing review. E-commerce and content channels should include a copyright-source check before listing goods or publishing content.
Third, distinguish “generation” from “copying”. True AI generation is new output formed from training; copying, collage and style-application to existing works are uses, and cannot be disguised as “AI-generated”.
Fourth, watch platform responsibility. Platforms providing generative tools should have complaint and takedown mechanisms to avoid becoming part of an infringement chain — see our algorithm filing service.
AI lowers the bar to content production, and the cost of infringement. Precisely for that reason, judicial evaluation is tightening. Controlling the copyright source is the most basic and most easily skipped step in AI content business.
For support on AI content copyright review, infringement risk assessment and criminal compliance, see our IP protection service and criminal compliance service.
This article is based on public information concerning the Beijing High People’s Court’s release on 22 April 2026 of its 2025 Typical Cases on Judicial Protection of New Quality Productive Forces. It is for general reference only and does not constitute legal advice.