Examining the critical difference between human imitation, AI deepfakes, and the legal concept of "substantial similarity" in music copyright.

The rise of Generative AI has blurred the line between respectful homage and illegal infringement. For decades, the music industry has wrestled with the legal challenge of the **soundalike**—a human vocalist hired to imitate a famous singer’s voice. Today, AI has perfected this imitation, creating a new challenge: the **deepfake**.

Understanding the current legal precedents regarding imitation is crucial, as courts will apply these decades-old standards to determine liability for new AI-generated music.

Deepfake vs. Soundalike: The Legal Distinction

The distinction between the two forms of imitation lies in the technology and the legal right they violate:

While the goal is the same—to evoke the famous artist—the law treats the direct, computer-generated replica differently from the human mimicry of a style. The **NO FAKES Act** specifically targets the AI deepfake, but the law already had tools for human soundalikes.

Protecting the Voice: The Right of Publicity Precedent

In the United States, an artist’s distinctive voice is protected not by federal Copyright Law, but primarily by state-level **Right of Publicity** laws, which prevent the unauthorized commercial use of an individual’s identity. Two landmark cases established this protection:

Midler v. Ford Motor Company (1988)

Bette Midler successfully sued Ford after the company hired a soundalike to imitate her distinctive voice for a Mercury Sable commercial, after Midler herself had refused the offer. The Ninth Circuit Court ruled that **"a voice is legally as distinctive and as personal as a face."** [1]

Waits v. Frito-Lay (1992)

Tom Waits won a suit against Frito-Lay for a Doritos commercial that used a soundalike to mimic his unique, gravelly vocal style. Waits was awarded $2.475 million in damages. The court emphasized that the imitation was designed to suggest his endorsement, constituting unfair competition and misappropriation of his identity. [1]

Protecting the Style: The Substantial Similarity Test

When an AI generates a song that is not a vocal deepfake but merely mimics the structural elements of a human artist (e.g., the chord progressions, instrumentation, and rhythmic patterns that define the "Taylor Swift sound"), the legal challenge falls back onto **Copyright Law** and the **Substantial Similarity** test.

The Standard for Infringement

To prove copyright infringement of a musical composition, a plaintiff must prove two things: [2]

1. **Access:** The defendant had access to the plaintiff's copyrighted work. (In the digital age, this is often easy to prove.)
2. **Substantial Similarity:** The defendant's work is substantially similar to the plaintiff's protected expression.

Courts rely on **musicologists** to dissect the works, filtering out "unprotectable elements" (common chord progressions, simple rhythms, or generic lyrical themes). What remains is the "protected expression," which must be copied to constitute actionable infringement. [3]

Style is Not Protectable

Crucially, **musical style** is considered an unprotectable *idea*, not a protectable *expression*, under the Copyright Act. [4]

• **Idea/Expression Dichotomy:** While a particular sequence of notes (the expression) is protected, the *genre* or *style* (the idea) is not. For example, the blues structure is a widely available idea.
• **The Challenge:** A human producer—or an AI—is legally allowed to create a new song in the style of, say, The Weeknd, as long as it does not copy the unique, protectable expression (e.g., a specific melody or lyrical phrase) of any existing copyrighted Weeknd song.

Legal Implications for Generative AI

The distinction between the soundalike and the deepfake has profound implications for how AI is regulated:

Conclusion

The legal framework for imitation is clear: an artist's signature voice is protected as part of their identity, whether it's mimicked by a human or cloned by a machine. However, the legal protection for an artist's *style* is narrow, making it difficult to sue an AI for merely mastering a particular genre or aesthetic. Future litigation will focus on whether AI-generated stylealikes cross the line into copying protectable musical expression, setting new precedents for how much originality copyright demands.

References & Further Reading

1. CNN interview: AI, deepfakes and the music industry (Discusses Midler and Waits precedents) - Joe Bennett.
2. What Musicians Should Know about Copyright | U.S. Copyright Office.
3. Uncleared Melody: Musicological Factors Considered in Copyright Infringement Cases - Cowan, Liebowitz & Latman.
4. AI and the Sound of Music | Yale Law Journal (Discusses Idea/Expression Dichotomy).
5. What's in a Name…or a Voice? Protecting Artists from Deepfakes in the Music Industry - Princeton Legal Journal.