A registration form now forces an AI authorship disclosure

The writer's problem

An author who used an AI tool to draft part of a manuscript, and now wants to register the finished book's copyright, faces a form that asks what a human actually wrote. On 16 March 2023 the U.S. Copyright Office published a statement of policy in the Federal Register telling applicants how to handle that question, prompted by what the notice describes as a rise in applications that did not disclose AI-generated material, some naming AI tools only in a title or acknowledgment.

What the documents show

The guidance states that applicants have 'a duty to disclose the inclusion of AI-generated content' and to describe, in the application's 'Author Created' field, only the portions a human actually contributed. AI-generated content that is more than minimal should be listed in the 'Material Excluded' section rather than claimed as authored, and an AI tool or its maker should never be named as an author or co-author. This builds on the Office's existing human-authorship standard: the Compendium of Copyright Office Practices already stated the Office will not register work 'produced by a machine or mere mechanical process' without human creative input, a rule this guidance now applies to generative AI specifically. The Office's own AI initiative overview places this guidance at the start of a broader inquiry that later produced reports on replicas, copyrightability and training.

The editorial choice

An editor helping an author complete a registration should treat the disclosure duty as procedural, not punitive: the guidance offers a template phrase, describing the human-authored 'selection, coordination, and arrangement' around AI-generated elements, rather than barring AI tools outright. Applicants unsure how to characterize a mixed work can, per the guidance, simply flag that AI was involved and let examiners follow up, rather than guessing at a disclaimer and risking an inaccurate registration.

What stays with the author

The guidance does not say who, if anyone, owns the AI-generated portion once it is excluded from the claim; it only says that portion cannot be registered as this applicant's authorship. Nor does it revisit registrations already granted before an author realized more should have been disclosed; the notice separately describes a supplementary registration process for that correction, but making it is the applicant's responsibility, not the Office's.

• Which specific passages, images or arrangements in the finished work were authored by a human, and which by a tool?
• Has a previously filed registration been checked against this disclosure duty before it is relied on in a dispute?
• Does more than de minimis AI content in this work call for exclusion language in the application?

The guidance answers a narrower question than authors sometimes expect: it governs what can be claimed on a registration form, not who is ultimately entitled to profit from an AI-assisted work.