Don Tomlinson wrote or co-wrote every song on this site and without generative AI involvement. Every sound recording on this site is the product of its professional original sound recording having been uploaded to an AI music site solely for purposes of changing the vocalist, updating the production, and sometimes adding another genre (with the melody changed only derivatively to accommodate the additional genre). Virtually all companies in the music licensing industry require any song they license to have been 100% human written. All underlying musical compositions on this site were 100% human written. For comparison and other purposes, the original sound recordings of almost all the songs on this site may be heard at www.solidsterlingmusic.com.

AI as a Demo Singer: The Legally (and Morally) Clean
Middle Ground Few Are Talking About

By Don Tomlinson
August 2026

                Author’s Note and Disclaimer: I live in two worlds. Over the past 40 years in my first world, I have written or cowritten more than 600 songs (see BMI) in a variety of genres (I’ve had indie cuts and cable network sync licensing and some major artist near misses). For that same period of time and in my second world, I have been an entertainment lawyer. I have an LL.M. in intellectual property law from the University of Houston Law Center and taught Entertainment Law there as an adjunct professor for ten years. My law practice has included representing songwriters, independent music publishers, independent record labels, a music-based cable television network, many other parts of the music industry, and a world of practice in entertainment law generally. The following article was prompted by my personal use of Suno, which has resulted in more than a thousand versions of the 600+ songs mentioned above. These thousand songs are based on original professional demos produced and owned by me and uploaded to Suno. They may be found on this site. For anyone interested in hearing the original professional demos of these songs (for comparison and other purposes), virtually all of them may be found at www.solidsterlingmusic.com. The disclaimer I need to add here is that, in the writing of this article, I collaborated with Co-Pilot on portions of it. That said, I am absolutely satisfied with the veracity of the information provided by Co-Pilot and contained in this writing. As the sole author of a number of music-themed law review articles, I feel entirely qualified to make this assertion.

           Artificial intelligence has forced the music industry into a polarized debate. On one side are those who fear AI‑based music will destabilize copyright, authorship, and creative livelihoods. On the other side are those who see music AI as a new frontier for composition, production, and sonic experimentation. But between these two extremes lies a quiet, legally and morally stable middle ground — one that has received almost no attention in the trade press or in legal commentaries. This article seeks to provide a basis for discussion.

              I’m writing here about the use of AI not to compose music but to perform music a human already has written. This workflow is not speculative. It is already happening (I’m doing it every day), and it is not nearly as legally complex as some other AI issues.

              In the main, new songs have two principal markets: 1) recordation paid for, produced, and distributed by a record label, the melody having been performed on the recording by an artist contractually tied to the record label and with the instrumentation performed by work-made-for-hire studio musicians; and 2) existing sound recordings licensed, mostly, by the film and television industries to be synchronized (“synced”) with visual elements of a motion picture or television program. There is a big difference between the two. In the case of a “cut” by an artist, the demo recording, which attracted the artist, etc., to the song in the first place, is used only to demonstrate the song, e.g., pitched by a music publishing company (or an independent songwriter or music plugger) to a record label, producer, or artist. In other words, the demo recording is not in any way distributed to the public.

              Some demos have become famous for their historical value. I remember seeing in a Music Row lawyer’s office many years ago the framed acetate disk of the demo of “Heartbreak Hotel,” sung by a Texas-based songwriter/demo singer named Glenn Reaves in the “anxious, dramatic style” that had been used on Elvis’s Sun Records releases. That acetate disk was used by Mae Axton to successfully pitch the song to Elvis and Colonel Parker. Axton, who with Tommy Durden had written the song in 1955, thought the song needed a more polished vocal than the one that had been supplied by cowriter Durden. It must have worked. So the story goes, Elvis listened to the demo about ten times straight and then agreed to record the song.

              In the case of the “sync” license, on the other hand, the demo recording is treated as a “master” and is used directly in the motion picture or television program. In this context, the entities licensing songs for direct use have a more problematic view of Suno-sung recordings; not necessarily because they think the melody then no longer is the human product of the songwriter but because there are obvious and lingering questions relating to other aspects of the resulting Suno recording, such as the controversies (and lawsuits) surrounding the myriad real voices and studio musician performances involved. Their concerns are not philosophical; they are contractual. They need clean chain‑of‑title. They need insurable rights. They need assurance that no part of the track was trained on infringing data or generated from someone else’s copyrighted work. Until these issues are ironed out by lawsuit settlement agreements and non-lawsuit contracting, most sync licensing companies are fearful of any Suno output. One hopes these issues will be resolved soon.

The Under-the-Radar Use Case: AI as a Demo Singer

              This article is about demo recording, not recordings to be pitched to sync-licensing companies. For many decades, songwriters and music publishers have hired demo singers to perform on sound recordings of underlying musical compositions (I’ve been doing it since the early 1990s, and we know it was done in 1955). These singers interpret the melody, adjust phrasing, and deliver a performance that helps the writer or music publisher pitch the song (generally singing the writer’s melody better than s/he could, which is certainly true in my case; otherwise, the writer would sing the demo and save the cost of hiring the demo singer). Also in this way, vocals may be sung by males, females, or both, and genre experimentations may be produced. No one rationally argues that a demo singer becomes a co‑author of the underlying musical composition by singing the demo. No one rationally argues that a demo singer’s melodic variations create new copyrightable works. The law is clear: the songwriter/music publisher owns the underlying musical composition; the demo singer simply has performed a version of it.

              Tools like Suno allow a songwriter or publisher to upload a studio-based demo recording and generate a new performance — perhaps in a different genre, with a different vocal timbre, or with stylistic variations that naturally arise.  When moving from, say, country to R&B, the melody will necessarily change derivatively to accommodate the difference in the genres. Neither the live demo singer nor Suno creates new authorship; they merely render a new and slightly different performance or interpretation of the melody that remains a part of the underlying musical composition. Quite importantly, derivative performances are squarely within the exclusive rights of the owner(s) of the underlying musical composition, usually the songwriter(s) and music publisher(s).

Derivative Works: The Legal Backbone

              Under 17 U.S.C. § 106, the author of an underlying musical composition holds the exclusive right to create derivative works. When an AI system generates a melody recognizably based on the songwriter’s original composition, it is not creating new authorship — it is producing a derivative work of the underlying melody. Federal courts have repeatedly held that melodic interpretation does not defeat derivation when the expressive identity of the original melody remains recognizable.

• In Bright Tunes Music Corp. v. Harrisongs Music, Ltd., 420 F. Supp. 177 (S.D.N.Y. 1976), the court found infringement even where the defendant argued the melody had been changed enough to avoid similarity. This is the famous case involving the pop tunes “He’s So Fine” and “My Sweet Lord” with the court finding that Beatle George Harrison had infringed the earlier melody when he penned “My Sweet Lord.” Substantial damages were awarded but Harrison was allowed to keep the copyright going forward.
• In Williams v. Gaye, 952 F.3d 1051 (9th Cir. 2020), the Ninth Circuit upheld liability where signature melodic elements persisted despite stylistic changes. The court held that “Got To Give It Up,” a number-one song written and performed by iconic pop star Marvin Gaye, was entitled to broad copyright protection because musical compositions are not confined to a narrow range of expression. The court ruled that the Gaye composition was infringed by “Blurred Lines,” the world’s best-selling single in 2013, authored by Pharrell Williams, Robin Thicke, and Clifford Harris, Jr. The court awarded substantial damages and a 50% royalty going forward.
• In Skidmore v. Led Zeppelin, 952 F.3d 1051 (9th Cir. 2020) (en banc), the court emphasized that protectable melodic expression — not surface‑level genre or style — is the core of similarity analysis. This case involved the allegation that the opening guitar riff in “Stairway To Heaven” infringed an earlier guitar instrumental titled “Taurus” (because it was an instrumental, the guitar played the entire melody). A jury found “Stairway To Heaven” not to be infringing of “Taurus.”

              These cases collectively establish a simple principle: If the melodic contour, identity, or expressive elements remain recognizable, the work is derivative — not new. This is precisely what happens when AI performs a melody the songwriter already composed.

Chain‑of‑Title: The Industry’s Real Concern

              When AI is used only to perform a melody the songwriter already owns (and using the songwriter’s lyric, which, together with melody, are normally the only protectible elements of a song), the chain‑of‑title is clean:

• The songwriter owns the underlying musical composition.
• The AI output is a derivative performance of that composition.
• The platform (in my case, Suno) grants commercial rights to the user.
• No rights in the underlying work are transferred to the platform (the only transferred right is the right to include the uploaded music in the AI database).

              This is legally identical to hiring a demo singer.

              AI‑performed demos meet all of these requirements when the songwriter/music publisher controls the underlying composition. In fact, AI demo performance can solve long‑standing industry problems:

• Quick gender‑swap demos for pitching.
• Fast genre conversions.
• Affordable demo production for independent writers.
• Enhanced vocal quality.

              None of these uses introduce authorship ambiguity. None create new copyrightable works. None threaten chain‑of‑title. They simply modernize the demo‑production process.

The Industry Needs to Catch Up

              The current discourse around AI music is dominated by fear of fully AI‑generated compositions. That fear is understandable — and justified. But it has overshadowed a practical, legally sound use that benefits songwriters, music publishers, record labels, producers, and artists. AI as a demo singer is not a threat to copyright. It is not a challenge to authorship. It is not a chain‑of‑title risk. It is not “AI‑generated music.” It is a performance tool. And performance tools always have been part of the songwriter’s toolkit.

Conclusion

              The music industry is overdue for a more nuanced conversation about AI, at least in this context. When songwriters/music publishers use AI to perform melodies they composed/own, the legal analysis is straightforward: the songwriter and/or music publisher owns the underlying work, the AI output is derivative, and the chain‑of‑title remains clean.

              This middle ground — legally conservative, technologically modern — may ultimately become one of the more valuable and least controversial uses of AI in recorded music. It is time we start talking about it. My proposal: Music publishers and record labels should draft an acceptance statement in line with the thesis of this article. Here is my first draft:

              “Vocal recordings uploaded to Suno for purposes of using Suno to create derivative vocal recordings are acceptable-use AI. When you submit a recording with a Suno vocal based on an uploaded non-AI vocal, please expressly indicate your compliance with the process stated in the previous sentence and be ready to share the original uploaded demo.”

              Music submitters should draft delivery statements in line with the thesis of this article. Here is my first draft:

              “The vocals contained in this demo recording are based on non-AI vocal recordings uploaded to Suno for purposes of using Suno to create derivative vocals. The original uploaded vocal recording is available on request.”

              Like the title of one of my songs, it’s “a good place to start.”