AI and Your Songs: What the U.S. Copyright Office Said in 2024

Introduction

If you are a producer who uses AI tools in any part of your creative process, for arrangement sketches, lyric drafts, production reference tracks, or direct generation of musical elements, the U.S. Copyright Office's 2024 report is the most important regulatory document in your field that you probably have not read.

The Office published Part 1 of its Report on Copyright and Artificial Intelligence on July 31, 2024. It addressed digital replicas, AI systems that realistically replicate a person's voice or appearance, and recommended federal legislation to address their unauthorized creation and distribution. Part 2, addressing the copyrightability of AI-generated outputs, followed in January 2025.

Together, the two reports established a legal framework that every producer using AI tools needs to understand. The framework can be stated simply: work entirely generated by AI cannot be copyrighted. Work that involves genuine human creative input can be. The line between those two categories is the place where producers need to pay attention.

---

The Report's Core Position on Authorship

The Copyright Office's 2024 analysis built on a principle that has been part of copyright law for a long time: human authorship is a "bedrock requirement" of copyright protection. The Office had applied this principle as far back as 2019, when it refused to register a copyright claim for an AI-created image in a case called Thaler v. Perlmutter. (That case ultimately reached the Supreme Court in 2026, which declined to intervene, leaving the human authorship requirement intact.)

What the 2024 report did was translate that general principle into specific, practical guidance for the AI music and creative output context. The guidance breaks down into several categories:

Purely AI-generated output cannot be copyrighted. If you enter a prompt into a music generation tool and use what comes out without significant human creative modification, the resulting track is in the public domain. Anyone can use it. You own nothing.

A prompt alone is not sufficient human authorship. As the Copyright Office explained, "a user's entering of prompts into an AI model alone, even if those prompts are highly detailed, does not provide sufficient human control over the resulting AI-generated expression and work to render the user an author." The reason: the same prompt can produce different outputs each time, meaning the creator does not have sufficient control over the expressive elements.

Using AI as a tool within a human creative process can be protected. If you write a melody, load it into an AI system to generate surrounding harmonic and rhythmic material, and then make creative decisions about what to keep, edit, and arrange, the human-authored elements of that process may be copyrightable. The AI-generated elements, taken on their own, are not.

Expressive inputs that shape AI output can create authorship. If you provide recordings, sketches, or other expressive material that the AI uses as a constraint, such that the resulting output is shaped by your specific human creative input in a way that goes beyond general prompting, there is an argument for authorship in the resulting work, at least in the portions reflecting your expressive choices.

---

The Digital Replica Recommendation

Part 1 of the report addressed a more urgent problem for working musicians: the unauthorized use of AI to replicate someone's voice. The Office found that existing law was inadequate to address this and recommended that Congress pass federal legislation specifically protecting individuals from unauthorized digital replicas of their voices and likenesses.

This recommendation matters practically for producers. Working with AI tools that use real artists' voices without their consent, even if the result sounds like a new performance rather than a direct copy, places producers in legally uncertain territory. The Office was clear that it viewed this as a harm requiring legislative remedy, and several state laws have already moved in this direction.

As a practical matter: AI tools that allow you to generate music "in the style of" a specific real artist, or to apply an artist's vocal characteristics to new recordings, are operating in an area where the legal framework is developing actively and may change.

---

What "Meaningful Human Authorship" Means for Producers

The phrase "meaningful human authorship" is the key concept for working producers. The Copyright Office has not defined a specific percentage of human input required or a precise threshold. It stated that the determination will be made on a case-by-case basis.

However, the report's examples suggest what meaningful human authorship looks like in practice:

• Writing the core melody, chord progression, or lyrical framework before using AI to fill in or expand
• Making deliberate creative decisions about what to keep, discard, or modify in AI output
• Using AI tools that allow control over specific creative elements (such as arrangement, instrumentation, timing)
• Modifying AI output through additional performance, production, or compositional work

What does not constitute meaningful human authorship, according to the report:

• Writing a detailed prompt and using whatever the AI produces
• Selecting among multiple AI outputs without making creative modifications
• Using AI to approximate a finished product from a general description

The practical implication for producers: document your creative process. If your work involves both AI tools and human creative decisions, the human decisions are what create copyright. Being able to demonstrate the nature of those decisions, through session notes, version histories, or other records, matters.

---

The Industry Context in 2024

The Copyright Office's report arrived in the middle of active litigation between major record labels and AI music generation platforms. Sony Music, Universal Music Group, and Warner Music Group had filed suit in 2024 against Suno and Udio, alleging that those companies had trained their models on copyrighted recordings without authorization.

The copyright question being litigated there was different from the authorship question addressed by the Copyright Office's report. The labels were arguing that training AI models on their recordings violated their copyrights. The Office's report addressed whether the outputs of AI models could themselves be copyrighted. Both questions matter, and both were in active development in 2024.

For producers, the practical result is a landscape in which the copyright status of AI-generated music, and of the tools used to generate it, was genuinely unsettled. The Copyright Office provided clarity on the output side. The training side remained in litigation.

---

A Plain-Language Summary for Working Producers

1. If you use AI to generate music without significant human creative input, you cannot copyright the result. Anyone can use it. This is both a risk (others can copy it) and a consideration (if you built your track on an AI foundation without modifying it, your ownership claim is limited).

2. Your creative decisions are what make a work yours. The prompt is not a creative decision sufficient for copyright purposes. The arrangement choices, the editorial decisions about what stays and what goes, the performances you capture or create, those are your creative contribution.

3. Voice simulation tools carry additional legal risk. Using AI to simulate a specific real artist's voice puts you in proximity to an area the Copyright Office explicitly flagged as requiring new legal protection. Proceed with caution.

4. Document your process. In a world where the line between human and AI contribution is contested, the ability to demonstrate your specific creative input is practically valuable.

The underlying principle is not new. Human craft is what creates value in music that is protectable and attributable. AI tools can assist the creative process, but they cannot substitute for the creative judgment that makes a work yours.

---

FAQ

Q: Can I copyright an AI-generated song? A: Not if it was generated purely by AI without meaningful human creative input. The U.S. Copyright Office's position, reinforced by court decisions and the 2024 report, is that works created entirely by AI, without sufficient human authorship in the expressive elements, are not eligible for copyright protection.

Q: Does using AI tools mean my music can't be copyrighted? A: No. Using AI as a tool within a human creative process does not automatically bar copyright protection. If you write core musical elements, make deliberate creative decisions about what AI output to use and how to modify it, and contribute meaningful human creative judgment, the resulting work (or at least the human-authored portions) may be protected.

Q: What is a digital replica under the 2024 Copyright Office report? A: The report defined digital replicas as AI-generated or digitally manipulated content that realistically depicts a real person, including their voice, in a way that is false or misleading. The Office recommended federal legislation to protect individuals from unauthorized digital replicas, particularly of their voices.

Q: If someone uses AI to generate music that sounds like my style, can I do anything about it? A: Style itself is not copyrightable under current law. However, if specific elements of your recordings are being used to train AI models without authorization, that is a different question, and one that the major-label lawsuits of 2024 addressed directly. The legal framework for this specific situation was still being developed as of 2024.

Q: Did the 2024 report change existing copyright law? A: No. The Copyright Office affirmed that existing copyright principles apply to AI-generated works. The report provided guidance on how those principles apply in AI contexts, but it did not require new legislation to address the copyrightability question. It did recommend new legislation for digital replicas specifically.

---