Major record labels and industry bodies have spent the summer rolling out a labeling framework to identify AI-generated music. The stated purpose is transparency for listeners. But the framework contains a structural gap that works in favor of the companies who drafted it.
On July 10, the RIAA, IFPI, the Recording Academy, SAG-AFTRA, and others announced voluntary track labels to distinguish between "AI-Generated" and "AI-Assisted" recordings. The distinction is straightforward in theory. A track qualifies as AI-Assisted rather than AI-Generated if "the recording was created substantially by humans and expresses human creativity" and "humans performed the lead vocal and primary instruments."
The critical phrase is "primary instruments." The system focuses on who performs the recording, not who wrote it. The labeling framework explicitly states it does not cover AI use in lyrics, composition, music videos, or cover art. That omission is significant.
The Performance Test
Under the current guidelines, a major label could use AI to generate an entire song. It could generate the melody, chord progression, arrangement, and backing track. As long as a signed artist records the lead vocal and performs the primary instruments, the track qualifies as AI-Assisted. If the AI contribution is limited to "some expressive elements," the song may not need the AI-Generated tag at all.
Apple Music has announced that it will begin displaying "Made With AI" labels to listeners by the end of 2026. The platform requires distributors to disclose AI use when "a material portion" of a track was created with AI. But Apple has not published a threshold for what counts as material.
Deezer has built its own AI detection system. It reported receiving roughly 90,000 fully AI-generated tracks per day in July 2026. Those tracks exceeded 50 percent of new uploads at peak times during June. Apple Music chief Oliver Schusser has said that approximately one-third of all new tracks on the platform are AI-generated. Yet those tracks account for just 0.5 percent of total streams.
Who Loses When the Labels Don't Have to Disclose
Session musicians work on a for-hire basis. They do not hold copyright in the recordings they contribute to. The major labels have publicly assured their signed artists and songwriters that no music will be ingested by AI without consent. But session players have no such leverage. Their performances can be replaced by AI-generated stems, and they have no recourse.
The economics are not subtle. Production costs for session musicians, string arrangements, and studio time run high. AI tools can generate orchestral arrangements in minutes. Labels have already started using AI outputs as demo references or final production elements for lower-budget releases. The workforce impact is concentrated among the players who have historically filled out recordings but never owned them.
Independent artists, meanwhile, face a different calculus. They do not have contracts with major vocalists. If they generate a song with AI, they cannot simply hand it to a celebrity to record and thereby qualify the track as human-made. The labeling system creates a class divide: those with access to major-label infrastructure can use AI invisibly, and those without are forced to disclose or risk removal.
The Disclosure Gap
The RIAA and IFPI have framed the system as a consumer transparency measure. But the framework is voluntary, and enforcement is left to whoever chooses to adopt it. The EU AI Act's Article 50, which took effect August 2, 2026, introduces binding transparency requirements for AI-generated audio. But its obligations fall primarily on AI providers and deployers of deepfakes. A label producing a track with AI-generated backing and a human vocalist may fall outside the strictest definitions.
Major labels have also proposed chart eligibility rules that would keep AI-Generated tracks off the Billboard 200 and equivalent charts worldwide. Universal, Sony, and Warner are among the companies calling for songs to be excluded unless they are "substantially human made." The definition of that phrase remains vague.
The structural advantage is clear. Labels control the means of production and the roster of artists who can humanize any track. Independent musicians and the session players who used to record those tracks are left with disclosure obligations that labels can route around.
Congresswoman Deborah Ross reintroduced the Protect Working Musicians Act in May 2026. The legislation would allow independent artists to collectively negotiate with streaming platforms and AI developers. Whether it passes, and whether it addresses the labeling gap, remains to be seen.
The industry's transparency push presents itself as a safeguard for human artistry. For major labels, it may function as something else: a competitive moat that protects their ability to use AI while stigmatizing competitors who lack their infrastructure.
