Cos'è: An analysis of the impact of generative AI models on the music industry: from platforms that produce complete songs from a text prompt, to legal battles over copyright, through to the pragmatic adaptation of industry professionals.
Suno and Udio: Complete Songs from a Prompt, in 2024
Until 2023, AI systems for music were primarily sound design tools or instrumental loop generators — useful but far from complete music production. 2024 changed this perception radically. Suno, a Cambridge (Massachusetts) startup founded by former Kensho researchers, made version 3 of its model available to the public in March 2024: by entering a short text prompt — style, mood, subject — the system generates complete tracks with vocals, lyrics, instrumentation and song structure (intro, verses, chorus, bridge) in seconds. The quality is sufficient to fool inattentive listening. Udio, launched almost simultaneously with a16z backing, offers similar capabilities with finer granular control over musical structure. Within a few months both platforms had reached millions of active users, changing the perception of what 'making music' means.
The 'Heart On My Sleeve' Case: The First Mainstream Scandal
In April 2023 an anonymous user on TikTok and Spotify published 'Heart On My Sleeve', a song apparently performed by Drake and The Weeknd, with convincing vocal styling from both artists. Within hours it had accumulated millions of streams on Spotify and views on TikTok before being removed for terms of service violation at Universal Music Group's request. The anonymous producer — using the pseudonym 'ghostwriter977' — had used AI models to clone the voices of the two artists, not directly protected by copyright in their pure form (in the US, copyright protects specific recordings, not the 'voice' as such). The case triggered immediate legal and industrial debate: Universal sent removal requests to Spotify, Apple Music and YouTube, and began pressuring AI platforms to block training on protected catalogs. The song, paradoxically, was stylistically credible — which made the situation even more destabilizing for the industry.
RIAA vs Suno and Udio: The Lawsuits That Will Define the Industry
In June 2024 the Recording Industry Association of America (RIAA), representing the major record labels (Universal, Sony, Warner), filed separate lawsuits against Suno and Udio in federal courts in Massachusetts and New York State. The charges are of copyright infringement on a massive scale: according to the RIAA, both platforms trained their models on protected music catalogs without licensing or compensation, and are capable of reproducing stylistic elements and sometimes recognizable sound fragments from the original songs. Suno and Udio's response was based on the fair use doctrine, arguing that training on protected works is a form of transformative learning comparable to human learning. This thesis is the same at the center of numerous other AI disputes in different sectors (New York Times vs OpenAI, Getty vs Stability AI). The outcome of the cases against Suno and Udio, expected by 2025-2026, will de facto define the legality of training on protected data across the entire generative AI ecosystem for creative content.
SAG-AFTRA, Hans Zimmer and the Adaptation of Professionals
Not all of the industry has positioned itself in frontal opposition to AI. In July 2024 SAG-AFTRA — the American union of actors and performers, which also includes voice actors and musicians — signed an agreement with several major studios to regulate the use of AI in productions: the terms include explicit performer consent for vocal cloning, compensation for the use of digital biometric data and protection against systematic replacement. This is a model of contractual coexistence, not a technological block. On the composers' side the situation is more nuanced. Hans Zimmer, interviewed in 2024, expressed concern about AI scoring not for principled reasons but practical ones: budgets for soundtracks are shrinking because producers purchase AI-generated music for secondary scenes and trailers at a fraction of the cost, reducing demand for mid-level composers. High-profile composers remain irreplaceable for premium projects; those in the middle market segment are most exposed. For independent music producers, conversely, Suno and tools like Udio or AI stem generation (separation and re-generation of individual tracks) are becoming normal workflow tools, used for rapid prototyping and variant generation.
The Open Question: Who Owns the Copyright of AI-Generated Music?
The United States Copyright Office established a clear principle in March 2023 for visual works (Zarya of the Dawn case): works generated entirely by AI are not copyrightable because they lack 'human authorship'. Music generated by Suno or Udio falls into the same category: whoever enters the prompt does not obtain copyright over the output. This creates a relevant commercial paradox: AI music generation platforms are building businesses on output that no one owns, freely distributable and impossible to protect as an intellectual asset. For creators who want to protect their hybrid productions — where AI contributed but human work is substantial in selection, arrangement and post-production — the Copyright Office has left open a grey area that courts will need to progressively delineate. In Europe the regulatory situation is even more uncertain: the DSM directive and the new AI Act do not explicitly address the copyright of AI-generated works with the clarity of the American system.
Link alla fonte originale
Official Suno platform for AI music generation. Useful for directly understanding system capabilities and updated terms of use.