Could Prompt Property Establish Ownership of the Knowledge and Intellect Behind Artifacts Made with AI?

Who owns the knowledge and intellect involved in an artifact made with AI? The question arises when a human supplies direction while a trained model supplies much of the expressive execution. This paper argues that ownership of a model, control of inference, and authorship of a prompt establish different possible claims; none alone establishes ownership of the knowledge underlying an output. Its technical premise is that parametric language generation is not a source-addressable database lookup. A generated answer does not automatically carry a verifiable genealogy of the human contributions and permissions that enabled it, although models can encode factual information and reproduce training material. The paper develops Prompt Property (PP) as a proposed account of attributable human direction: a claim rooted in the instructions and decisions a person actually contributed. PP may support assessment of a qualifying contribution or protection of an instruction under an applicable regime. It cannot, merely through a prompt record, turn the model's learned substrate into the prompter's property. The argument distinguishes human intellectual creation from current legal title, production provenance from training provenance, and authorship-based claims from rights with other qualifying bases. Category differences delimit the argument rather than replace its central question. The paper concludes that the defensible object of a PP claim is a specific attributable contribution, not the total knowledge and intellect behind an AI-mediated artifact. This is a conceptual and normative proposal, not an empirical finding or a new legal entitlement.