On 18 May 2026 media reported that the Shanghai Huangpu District People’s Court decided a copyright dispute arising from AI prompts, holding that a simple, list-style set of prompts is not a work and dismissing the claim of a cultural company as plaintiff. The case shifts the legal focus from AI-generated output to the starting point of the creative chain.
The dispute and its focus
According to media reports, the plaintiff claimed copyright in a set of prompts used to generate AI content and accused others of infringement. The focus therefore fell on whether a prompt itself can be a “work” in the sense of copyright law.
The question is unusual because in AI settings “creation” is usually understood as the image, text or video the model outputs, while a prompt is often seen as the means to that result. When the means is claimed as a right in itself, the law must answer whether it meets the elements of a work.
Why a prompt is not a work
Copyright protects expressions that are original and capable of being fixed in some form. Prompts are typically made up of instructions, parameters and style descriptions whose function is to issue requirements and steer the output. Where they are merely a simple list of requirements, lacking individualised selection and arrangement, they struggle to constitute original expression.
The ruling lands exactly here: a simple, list-style prompt is a statement of instructions, closer to an “idea” or a “method” than to protected “expression”. This finding does not deny the technical or commercial value of prompts; it says only that they fall outside the reach of copyright.
Applying the expression/idea dichotomy in AI settings
“Protect expression, not ideas” is a basic divide in copyright. AI settings make it newly tense: users invest great effort in refining prompts and intuitively regard this as “creation”, yet structurally a prompt drives a generation process and its value lies in “how to obtain the result” — closer to a method or a rule.
A distinction is needed between “a prompt” and “a work containing prompts”. Where prompts are arranged into an original document, textbook or process manual, the whole may be protected; a scattered, functional set of instructions usually is not protected on its own.
What it means for companies protecting AI outputs
First, put the focus of protection back on the output itself. AI-generated content that is original can be claimed through copyright, and companies should keep records of the generation process to show human intellectual input — see the AI IP protection guide.
Second, protect prompts as trade secrets. Where a prompt has commercial value and is protected by confidentiality measures, it can be protected as a trade secret, a route that fits its “method” nature better than copyright — see the AI criminal compliance guide.
Third, allocate ownership by contract. Agreeing in writing with developers and partners on ownership of prompts, workflows and outputs is more reliable than litigating later — see the AI commercial contracts guide.
Fourth, claim compilation and systematisation separately. Where prompts are organised into a systematised template library or process standard, a combined protection through compilation works or trade secrets may be considered.
For support on rights strategy for AI outputs, trade-secret protection and contractual arrangements, see our IP Protection and Commercial Contracts services.
This article is based on public media reporting on 18 May 2026 of the relevant case before the Shanghai Huangpu District People’s Court. It is for reference only and does not constitute legal advice.