How record labels and publishers are fighting back against unauthorized AI training and the rise of the "licensed ecosystem."

While the NO FAKES Act focuses on the protection of an individual's identity (voice/likeness), the primary legal struggle for the music industry is the massive, unauthorized use of their existing catalog—billions of songs—to build generative AI models. This conflict centers on two key areas: the legal right to use music for **training data** and the question of **who owns the output**.

The Training Data Dispute: Fair Use vs. Theft

The foundation of every powerful AI music generator (like Suno, Udio, or Jukebox) is a vast library of music used to teach the model about melody, rhythm, genre, and harmony. The dispute boils down to the doctrine of **Fair Use** under U.S. Copyright Law.

The AI Developers’ Argument (Fair Use)

AI companies argue that the act of copying billions of songs to train a model is a **transformative fair use** of the material. [1]

• **Transformative Use:** The training process involves converting the copyrighted works into mathematical representations (weights and tokens), which is fundamentally different from the original artistic expression.
• **New Creative Tool:** The output (a newly generated song) does not compete directly with the input (the original training songs), as the output is a new product—a creative tool.

The Rightsholders’ Argument (Infringement)

Major record labels (Universal Music Group, Sony Music) and publishers argue that this is massive, willful copyright infringement that undermines their ability to control and profit from their work. [2]

• **Scale of Copying:** Making millions of copies of every song to ingest into a model is unauthorized reproduction and a derivative work—the core of copyright infringement.
• **Market Harm:** If an AI can generate a perfect functional equivalent of a songwriter’s style instantly, it directly harms the market for their future music and their licensing potential.

Why Pure AI Music is Denied U.S. Copyright

While the input (the training data) is protected, the U.S. Copyright Office has been clear: the output of a purely generative AI model is **not** eligible for copyright protection.

The "Human Authorship" Requirement

The U.S. Copyright Office requires a creative work to be the product of **"human authorship"** to qualify for protection. [4]

• If a human merely gives a text prompt (e.g., "Write a pop song about a breakup") and the AI does the rest, the resulting music lacks the required human creative input and is placed into the **public domain**.
• This has pushed AI platforms to adopt **human-AI co-creation** models where users must demonstrate meaningful creative control (editing, arranging, remixing) to secure copyright for the final track.

This denial of copyright is a double-edged sword: it prevents AI developers from securing protection for fully automated outputs, but it also means those outputs can be freely used by anyone, potentially flooding the market.

The UMG Strategy: Litigation and Licensing

Universal Music Group (UMG), the world’s largest music company, has taken the lead in establishing the industry’s response, moving from pure opposition to a highly structured model of **"The Walled Garden."**

Their strategy operates on two parallel tracks:

Track 1: Establishing Legal Authority (The Lawsuits)

UMG used litigation against key AI companies to force them to the negotiating table. This established the legal precedent that they will fiercely defend their intellectual property rights against unauthorized training.

Track 2: The Pivot to Licensed Collaboration

The most significant trend is the transition from fighting AI to partnering with licensed AI models. The landmark partnership between UMG and Udio is the clearest example: [5]

• The partnership aims to launch a new generation of subscription-based AI tools.
• The AI models will be trained exclusively on **authorized and licensed music**.
• This creates a "safe" ecosystem where artists are compensated for both the use of their music as **training data** and for any successful commercial outputs generated by the AI.

Goal: The music industry's aim is to ensure that AI innovation occurs within a healthy, commercially licensed framework, ensuring rightsholders are compensated, rather than allowing the AI economy to be built on an unlicensed foundation.

The Future: New Compensation Models

Traditional royalty splits are becoming outdated when a single AI prompt can generate a song. The industry is exploring new, more sophisticated models:

The Attribution Share Model

This proposed system would compensate original artists and rightsholders proportionally to how often their creative assets (specific sound recordings, chord progressions, stylistic elements) are detected as having **influenced** a new AI-generated output. [6]

• **Pro:** Rewards the value of the foundational training data, even if the final output is deemed "new."
• **Con:** Requires highly sophisticated (and expensive) AI detection technology to accurately trace influence across billions of data points.

Blockchain and Metadata Solutions

The long-term goal is to use blockchain and other secure technologies to create immutable, hyper-detailed metadata embedded in every piece of music. This metadata would clearly state ownership and licensing terms, making it easier for AI models to legally ingest and track usage for proper payment.

Conclusion

The legal fight over AI music is quickly evolving from a battle over whether AI can use music to a negotiation over **how** it can use music. Driven by landmark litigation and the powerful pivot by groups like UMG, the trend is moving toward the creation of licensed, compensated AI models. The success of this transition will determine whether generative AI becomes a destructive or creative force for human artists.

References & Further Reading

1. Fair Use Index - U.S. Copyright Office.
2. RIAA Policy & Advocacy: Artificial Intelligence.
3. Major Music Labels Sue AI Company Suno Over Copyright Infringement - The Hollywood Reporter.
4. Copyright Registration Guidance: Works Containing AI-Generated Material - U.S. Copyright Office.
5. Universal Music Group strikes deal with Udio AI music generator - Music Business Worldwide.
6. Compensation Models for the Use of Music in AI Training - WIPO Magazine.