Beijing Internet Court Decision Upheld: Prove Creative Input for AI Image Copyright

Beijing Internet Court says AI images can earn copyright if creators document prompts, iterations, selections, edits, and timestamps. No records, no rights; a case was lost.

Categorized in: AI News Creatives
Published on: Sep 17, 2025
Beijing Internet Court Decision Upheld: Prove Creative Input for AI Image Copyright

Beijing Internet Court: Show Your Creative Process to Claim Copyright on AI Images

On September 16, 2025, the Beijing Internet Court confirmed that AI-generated images can get copyright protection-if the creator proves real creative effort. You must show your thinking, the prompts you used, how you iterated, what you selected, and what you edited. No process, no protection.

This matters for creatives who use Midjourney, Stable Diffusion, Firefly, or similar tools. The court wants proof that your hand guided the output, not just an image that happened to come out of a model.

The Case at a Glance

A content creator, Zhou, claimed rights in an AI-generated "Cat Crystal Diamond Pendant" image shared during a business partnership. The defendant allegedly used the image across platforms without permission in October 2023 and March 2024.

Zhou provided only after-the-fact recreations using AI during litigation, not the original generation records. The court rejected this as insufficient because it didn't reflect the original prompts, iterations, or selection and modification steps.

Outcome: The court ruled for the defendant, and the decision was upheld on appeal.

What the Court Expects as Proof

  • Explain your creative intent and decisions.
  • Provide original input prompts, seeds, and key settings (model/version, parameters).
  • Show iteration history: prompt tweaks, variation rounds, and upscales.
  • Include selection records: which generations you kept vs. discarded and why.
  • Document edits: post-processing steps, layers, and revision timelines.
  • Back it up with timestamps, file hashes, export logs, or platform history.

Judge Wang Yanjie's Key Points

  • The burden of proof mirrors traditional works: whoever asserts must provide evidence.
  • Because AI needs less manual effort, originality is judged case by case-your documented process is the difference-maker.

Protect Your AI Image Rights: A Practical Checklist

  • Turn on and retain platform histories. Export prompt logs, seeds, and settings after key milestones.
  • Save every iteration batch with timestamps. Keep both chosen and rejected outputs.
  • Record edits in your editor (layered files, adjustment logs, plugin history).
  • Use file hashes and cloud backups to show an unbroken chain of creation.
  • Avoid "recreating" later. Courts want original records, not simulations.
  • If collaborating, document who did what (prompt authorship, curation, editing) and agree on credits upfront.
  • When publishing, keep dated proofs (posts, emails, project tickets) linked to the exact files you claim.

Industry and Policy Context

The court also encouraged stronger model traceability and implementation of rules for identifying AI-generated content. For reference, see China's deep synthesis regulation overview here: Deep Synthesis Provisions.

Prior Signal from the Courts

In late 2023, the Beijing Internet Court recognized copyright protection for AI-generated images in Li v. Liu. The latest ruling doesn't roll that back-it clarifies the proof you need.

Bottom Line

If you want to claim rights in AI art, build a trail. Prompts, seeds, iterations, selections, edits, and timestamps-treat them like your brushstrokes.

If you want to improve your prompting workflow and documentation, explore practical resources here: Prompt Engineering Guides.