The Copyright Office says digital replicas need federal law

The writer's problem

A ghostwriter, biographer or memoir collaborator increasingly works alongside tools that can rebuild a real person's voice or face well enough to pass as authentic. The first installment of the U.S. Copyright Office's Report on Copyright and Artificial Intelligence, published 31 July 2024, takes up this problem: what happens when a digital replica, a video, image or audio recording that realistically but falsely depicts a real person, is made and distributed without that person's permission. The report followed an August 2023 notice of inquiry that drew, per the Office's own AI initiative page, extensive comment on whether existing law already covers this harm, or a new federal law is needed.

What the documents show

The Office states plainly that it concluded 'a new law is needed,' finding that state privacy and publicity laws, the Copyright Act, the FTC Act, the Communications Act and the Lanham Act do not together provide a reliable nationwide remedy. Framed as recommendations rather than a binding rule, the statute should cover realistic replicas 'whether generated by AI or otherwise,' should protect every individual rather than only celebrities, should attach liability to distributing a replica rather than to its creation alone, and should require actual knowledge that a depiction is both a replica and unauthorized. The Office also recommends licensing rather than outright assignment of the right, a safe harbor for online services, and no full federal preemption of state laws. Separately, it states the Office does not 'recommend including style in the coverage of new legislation at this time,' a narrower conclusion than the headline recommendation on replicas.

The editorial choice

For an author or estate deciding how to respond to an unauthorized replica today, the practical fact is that Part 1 is a recommendation, not a law on the books. An editor advising a client whose voice or image has been cloned should route the question to existing state right-of-publicity statutes and the other federal frameworks the report surveys, rather than assuming a new federal cause of action already exists. Treat the report as a preview of where Congress may go, while telling readers plainly that its recommendations remain proposals.

What stays with the author

The report is explicit about what it does not resolve. It says nothing about whether an AI-generated work can itself be copyrighted, since that waits for Part 2, and nothing about whether training a model on copyrighted works is lawful, the subject of Part 3. An author's own diligence about a collaborator's likeness in promotion or narration is not answered here; it stays a contractual and consent question the report does not settle.

• Does the jurisdiction where the affected person lives already have a right-of-publicity statute covering this use?
• Is the disputed material being distributed or merely created, given the report's proposed distinction?
• Would treating a recommendation as settled law mislead a client about their actual legal options?

Part 1 is one section of a report still being written in installments; its value to a working editor is naming the gap in existing law precisely, not supplying a remedy that does not yet exist.