When a business buys an AI-enabled software platform and feeds it prompts and data, the platform produces outputs: text, analytics, images, code. Who owns those outputs, and are they protected by copyright at all? The instinctive answer, that whoever paid for the work owns it, is wrong in law. Copyright statutes in India, the United Kingdom and the United States are built around a human author, and an output generated by a machine with no human creator sits awkwardly, or entirely outside, that framework. The practical consequence is that ownership of AI-generated material cannot be presumed. It must be allocated expressly by contract. This explainer sets out the statutory position in the three jurisdictions, the case law on machine authorship, and the drafting approach that follows.
The Threshold Question: Does an AI Output Attract Copyright at All?
Before ownership can be argued about, there has to be a right to own. Copyright subsists only in a "work" with an author, and the recurring difficulty across jurisdictions is that copyright law was written on the assumption that authors are human. Where a human directs the process, prompting, selecting, arranging and validating the output, the human authorship requirement can generally be satisfied and the output is protectable in the ordinary way. Where the output is produced autonomously by the machine with no meaningful human creative input, the protection is uncertain or absent. Most jurisdictions have not legislated specifically for the "protectability" and "ownership" of AI-generated outputs, leaving the question to be resolved either by straining existing statutory definitions or by contract.