Wuhan court: an AI micro-drama is a copyrightable film

A district court in Wuhan has held that a micro-drama made mostly with generative AI tools is a copyrightable work, and counted the cost of the tokens burned generating it when it set damages. The Hubei copyright protection and service site carried Hubei Daily's report of the ruling on September 21, 2026, and People's Daily Online's Hubei channel followed on September 22 with the judge's explanation.

The facts, as the two reports give them: a professional micro-drama company, identified only as Company A, made a 47-episode drama of about an hour in total in early 2026, using generative AI tools. It filed the drama with the National Radio and Television Administration and released it on the Hongguo Short Drama app and on WeChat Channels. One day later, Company B copied all 47 episodes to its own WeChat Channels account under a changed title and ran commercial advertising against them. Company A sued in the Jiang'an District People's Court for infringement of its right of communication through an information network; Hubei Daily's report describes it as the province's first concluded case of its kind.

The question the court had to answer was whether a drama whose images and voice track were largely machine-generated is a work at all. Its answer turned on the production record rather than the output. According to both reports, the process ran through script planning and copywriting, then storyboard prompt-setting, selection of character and scene material, screening and discarding of generated clips that did not fit, and finally picture editing, detail correction and matching of audio and subtitles. The company's staff took part in all of it. The court found the finished drama was not the product of a single automated pass, that the staff steered the output toward a preset creative intention using their own aesthetic judgement, and that they had "complete foreseeability and substantive control" over the final expression. The AI tool, in the court's framing, was only the instrument; the drama is an audiovisual work.

Judge Yi Jingwen put the test in the same terms in People's Daily Online's account: the core requirement is a human intellectual achievement with originality, and the assessment cannot look only at whether the final picture was generated by AI but at the depth of human participation in the process.

On damages, neither side discharged the burden of proving actual loss, the infringer's profit, or a licensing fee for the work, so the court fell back on the statutory factors: the drama's runtime, its reach, the window in which it was popular, how long the infringement lasted, and the defendant's fault. It then did something less usual. Weighing creative cost, and taking account of how AIGC production actually works, it counted the computing cost of token consumption and the cost of licensing commercial AI tools. The award was RMB 20,000 (about $2,800), covering the economic loss and reasonable expenses. Neither party appealed, and the judgment is final.

Two limits on its weight. It is a first-instance ruling from a district court, reported by two Hubei outlets rather than published as a judgment text, and the parties are identified only as Company A and Company B and the drama's title is a pseudonym, so nobody outside the case can check the record it turned on.

What travels regardless is the shape of the reasoning: in this court's view, what protects AI-assisted work is the documented human chain around the model. If you are making micro-dramas or any AI-assisted series, the file that wins this kind of case is the one that already exists in your workspace: the script, the prompt revisions, the keep-and-kill list, the edit timeline. The token bill belongs in it too, because this court counted it as a creative cost when it set damages.

The next procedural step is the one that did not happen: no appeal was filed, so the reasoning stands as a district court's, and the point to watch is whether a higher court adopts it in a case with named parties.