A Chinese Court Just Made AI Token Costs a Legal Factor in Copyright

For decades, copyright law rested on a simple premise: human creativity has value. A court in Wuhan, China just added a comma to that sentence, and the clause that follows changes everything.

Token usage. API licensing fees. The cost of inference compute. These are not terms you would normally find in a copyright damages ruling. But in a decision handed down this month, the Wuhan court factored exactly these AI production costs into its calculation for the first time anywhere. The message is clear: in China, AI compute is becoming a form of intellectual labor that the law can quantify.

And the rest of the world is still figuring out how to value content nobody "created."

An AI-generated drama, copied overnight

The case centers on a one-hour short drama produced using AI tools in early 2026. A company used commercial AI platforms to generate the script, design prompts, select outputs, and edit the final version before publishing it on platforms like WeChat.

One day later, another company copied the work, gave it a new title, and ran advertisements against it.

The plaintiff sued for copyright infringement. The court had to answer two novel questions: could an AI-generated work qualify for copyright protection at all, and if so, how do you calculate damages when the "creative" input was a text prompt and a compute bill?

The tool, not the creator

The court's first ruling was that the drama qualified as a protectable audiovisual work. The key reasoning: human employees had contributed at every stage. They wrote the script, designed the prompts, selected from AI outputs, and performed final editing. The AI was just a tool (a sophisticated one, but legally no different from a camera or editing software.

This extends a 2023 Beijing precedent that first recognized copyright in an AI-generated image. That earlier ruling established that AI-assisted works could be protected if human authorship was demonstrable. The Wuhan decision goes further by applying the same logic to a longer-form, commercially distributed AI-generated work.

Token costs as damages

Where the ruling breaks new ground is in its damages calculation. Beyond traditional factors like runtime, distribution reach, and infringement duration, the court explicitly weighed the cost of AI tokens used in production and the licensing fees of AI tools. These are production expenses in the AI era, and the court treated them as such.

The total award was 20,000 RMB (about $2,900). The amount itself is modest, but the principle is not. By recognizing AI compute costs as a compensable factor, the court has created a framework that could scale dramatically: the more sophisticated the AI production, the higher the token cost, and therefore the higher the potential damages.

The court also recommended that creators keep detailed records: scripts, prompt drafts, project files to substantiate their human contribution. This is practical advice born from a legal reality: if you want to claim copyright in AI-generated work, you need to prove the human was in the loop.

What this means beyond China

China has been aggressively building out its legal framework for AI-generated content. The 2023 image case, the 2024 generative AI regulations, and now this ruling form a coherent trajectory: China is creating a legal architecture where AI-assisted works have clear protection, and the costs of AI production are recognized as real economic inputs.

Other jurisdictions are far less advanced. The US Copyright Office has taken a more restrictive stance, requiring substantial human authorship and denying registration for purely AI-generated works. The EU's AI Act addresses risk but not copyright. The UK has consulted on reform without legislating.

The Wuhan ruling does not bind courts outside China. But it provides a concrete legal model and a glimpse of a future where "how many tokens did this take?" becomes a standard question in copyright litigation.

The unresolved questions

The ruling leaves several questions open. What happens when AI does most of the creative work and the human contribution is minimal? How do you value prompt engineering against traditional authorship? And if token costs are production expenses, should they also be deductible for tax purposes, or treated as capital investments in an AI training pipeline?

These are questions the market will answer before the courts do. Every AI-generated commercial work produced today is a test case for the legal framework of tomorrow.

For now, one thing is certain: a court in Wuhan has ruled that the cost of running AI models is a real, quantifiable form of labor. And in doing so, it gave the global AI industry something it has been lacking: a legal precedent for the economics of creation.

Sources

• The Decoder. "A Wuhan court just made AI production costs a legal factor in copyright infringement cases." September 28, 2026.
• The National Law Review. "Wuhan Court Considers AI Token Costs in Calculating Damages for Copyright Infringement of AI-Generated Work." September 24, 2026.
• Sudovo-Yuridichna Gazeta (Ukraine). "In China, a court recognized copyright for an AI series and considered token costs." September 26, 2026.