A digital replica right is a recommendation, not yet law

The system

A convincing fake of a singer's voice is now cheap to make and hard to catch before it spreads. The Copyright and Artificial Intelligence, Part 1: Digital Replicas report was published by the Register of Copyrights at the U.S. Copyright Office on 31 July 2024, the first installment of a multi-part study opened after an August 2023 notice of inquiry. It addresses one question: what the law should do about digital technology that realistically recreates a specific person's voice or image without consent. The Office frames this as a gap in existing law, and its AI initiative page situates Part 1 alongside later parts on AI-output copyrightability and training data, so a reader should not mistake this report for the Office's full position on generative music.

What the documents establish

The report states its conclusion plainly: the Office recommends that Congress create a new federal right protecting individuals, during their lifetimes, from the knowing distribution of an unauthorized digital replica of their voice or image. The right would be licensable but not assignable, carry damages and injunctive relief, apply ordinary secondary-liability rules with a conditioned safe harbor for online service providers, and include First Amendment accommodations. This is a recommendation, not a rule already in force; only Congress can create it. The report declines to extend the same protection to a performer's style alone, judging existing copyright and unfair-competition law adequate there. The AI initiative page confirms the publication date and shows Part 1 sits ahead of Part 2 (copyrightability, January 2025) and Part 3 (training, pre-published May 2025).

Craft and rights

For a producer or label, capability and rights point in different directions. A voice model can generate a plausible vocal take in a target artist's timbre in minutes; what remains a human decision is whether that artist agreed to it. The proposed right would matter most once a track is distributed with a replicated voice the performer did not authorize, since it targets knowing distribution rather than private experimentation. Until Congress acts, artists rely on state right-of-publicity laws the report calls inconsistent and hard to enforce across state lines. This is an editorial read of the gap: a session singer has less certainty today than a songwriter whose composition is copied outright, because voice and likeness sit outside standard copyright registration.

Outcomes and open questions

No federal digital-replica statute has been enacted from this recommendation, and the Office notes state laws would not be fully preempted even if one passes. Watch whether Congress introduces legislation tracking these contours, and whether Part 3's training-data conclusions get treated, wrongly, as resolving questions Part 1 kept separate.

• Does a release involve a voice or likeness the performer did not explicitly authorize for that use?
• Would today's contract language cover a synthetic replica made after the artist recorded it?
• If a replica surfaces without consent, what remedy exists before any federal right is enacted?

Read narrowly, Part 1 is a policy recommendation about one specific harm, not a verdict on generative music generally, and the distinction matters for anyone deciding how much protection today's law actually offers.