Executive Overview
As artificial intelligence platforms rapidly infiltrate the creative economies, a foundational battle over the economic anatomy of music is reaching a boiling point. The Independent Music Publishers International Forum (IMPF) and the international digital licensing hub IMPEL have formally unveiled a comprehensive, cross-border licensing framework designed to govern how generative AI companies interact with musical property. At the core of this aggressive new strategy is a single, uncompromising mandate: the composition—the underlying song, lyrics, and melody—must be valued equally to the sound recording, effectively dismantling decades of structural inequality established during the streaming era.
For generations, the commercial distribution of music has operated under a deep financial imbalance. In the traditional streaming market, master recordings (the sound recordings owned predominantly by major record labels) have consistently commanded the lion’s share of royalty revenues, leaving songwriters and music publishers to fight over a historically meager slice of the pie. Now, as tech conglomerates race to train massive large language and audio generation models on vast catalogues of copyrighted music, independent publishers are declaring that history cannot be allowed to repeat itself.
The newly published framework asserts that because songs, lyrics, and melodies are the absolute bedrock of AI training, prompting, and final generation, the economic return must reflect that reality. Beyond the headline demand for equal valuation, the framework emphasizes absolute operational transparency, clear delineation between training and output revenues, and a resolute refusal to accept licensing models that treat composition rights as an afterthought. With major record labels already cutting deals with AI startups and tech giants—frequently prioritizing master-heavy ecosystems like remix and voice-cloning tools—the independent sector’s intervention injects a vital, protective counterweight into an increasingly volatile marketplace.
Detailed Chronology: The Road to the IMPF/IMPEL Framework
The launch of the IMPF-IMPEL framework does not happen in a vacuum; it is the culmination of years of mounting tension between the technology sector and the global independent publishing community. Understanding how the industry arrived at this critical juncture requires examining the timeline of digital disruption, copyright evasion, and the shifting dynamics of music licensing.
Phase 1: The Streaming Precedent and the Original Sin
When digital streaming disrupted the physical and download economies in the late 2000s and 2010s, the music industry was forced to construct entirely new licensing mechanisms on the fly. Because major record labels controlled the master rights—and often held significant minority equity stakes in early streaming services—they leveraged their market dominance to secure favorable royalty splits. Consequently, sound recordings captured the vast majority of streaming revenue, while the mechanical and performance royalties allocated to publishers and songwriters lagged far behind.
For years, independent publishers fought pitched battles via collecting societies to incrementally close this gap. However, the foundational revenue split remained skewed heavily toward the master side. When generative AI began migrating from academic curiosities to commercial products capable of synthesizing human-sounding vocals and instrumentation, independent publishers immediately recognized the existential threat of carrying those outdated streaming assumptions into the machine learning age.
Phase 2: The Wild West of Unlicensed AI Training
Between 2022 and 2024, the generative AI boom accelerated dramatically. Tech companies—ranging from heavily funded Silicon Valley startups to multinational conglomerates—scraped millions of copyrighted sound recordings, musical compositions, and lyrics from the internet to train their foundational models. Crucially, much of this harvesting occurred without the explicit permission of or compensation to rightsholders.
As copyright infringement lawsuits began piling up in courts worldwide, technology platforms realized they needed to regularize their operations. They began approaching the music industry to negotiate retroactive settlements and forward-looking licensing agreements. However, these negotiations were largely spearheaded by major record labels and major publishers. Because major music corporations house both record labels and publishing arms under single corporate umbrellas, critics have long suspected an internal conflict of interest: major conglomerates often favor directing capital toward their record labels, where artists typically receive a fraction of the royalties, rather than their publishing arms, where standard industry practice mandates a 50/50 split of net receipts directly with the songwriter.
Phase 3: The Independent Counter-Offensive
Recognizing that initial AI licensing agreements threatened to lock in another generation of publishing undervaluation, IMPF and IMPEL mobilized. Working in tandem with global songwriter advocacy groups, the organizations began drafting a unified industry response.
The resulting framework is designed to serve as a practical playbook and a moral red line. It arrives precisely as AI companies are finalizing commercial rollouts, attempting to normalize models that treat the song as a secondary input. By issuing this framework, the independent publishing community has shifted from a defensive posture—reacting to deals cut behind closed doors by major players—to an offensive strategy, setting explicit terms of engagement for any tech firm wishing to access their repertoires.
Supporting Context & Metrics: The Economics of AI Music Training
To understand why the IMPF and IMPEL framework is so fiercely protective of the composition, one must examine the technical and economic mechanics of generative artificial intelligence as it applies to music.
The Dual-Stream Reality of AI Music Products
Unlike traditional streaming, where money flows through a single primary channel (per-stream consumption), generative AI products monetize music across multiple distinct touchpoints. The IMPF/IMPEL framework highlights that any comprehensive licensing deal must account for at least two primary income streams:
- Model Training and Ingestion: This involves the massive upfront ingestion of copyrighted musical works—both compositions and master recordings—to teach neural networks the patterns of human music, chord progressions, harmonic structures, and lyrical syntax.
- Outputs, Exploitation, and Interactive Tools: This encompasses the generation of new works, prompts, interactive remix tools, mashup engines, and voice-cloning technologies that utilize existing tracks or specific artist personas in their final consumer-facing outputs.
Each of these streams carries unique valuation challenges. For instance, in the training phase, how is the value of an individual song calculated? Does an AI company divide its training payment equally across an entire catalogue, or does it deploy data-driven attribution models to trace precisely how often a specific songwriter’s harmonic choices or lyrical styles influenced the model’s weights? Furthermore, many early AI deals include lump-sum legal settlements meant to compensate rightsholders for past infringement—monies that must be carefully segregated from ongoing operational royalties and distributed fairly down to the individual creator.
The Conflict of Major Conglomerates vs. Independent Publishers
The debate over valuation is further complicated by the structural differences between major music corporations and independent publishers. Major publishers (such as Universal Music Publishing Group, Sony Music Publishing, and Warner Chappell) operate alongside sister record labels within massive corporate ecosystems.
While major executives and publishers vigorously deny that corporate internal dynamics influence external licensing negotiations, independent publishers remain deeply skeptical. In a traditional recording contract, a signed artist is typically bound by recoupable costs and royalty rates that net the artist a minority share of the earnings. Conversely, in a publishing contract, the statutory and contractual splits overwhelmingly require publishers to pass the vast majority of royalties straight through to the songwriter. Therefore, from a purely financial optimization standpoint, a major conglomerate theoretically benefits more when a larger percentage of licensing revenue is funneled through the record label rather than the publishing division.
Independent publishers, who do not operate massive sound-recording distribution arms to offset these disparities, view the composition as their lifeblood. For them, allowing AI platforms to devalue the song is not merely a bad business deal—it is an existential threat to the economic viability of songwriting as a profession.
Official Statements and Industry Perspectives
The urgency and conviction driving the new framework are echoed by the leaders spearheading the initiative.
Annette Barrett, President of IMPF, emphasized the vital role of independent publishers in protecting the creative ecosystem during a period of profound technological disruption:
"IMPF exists to give independent publishers and the songwriters we represent a collective voice, a mission that has never mattered more, as our industry navigates disruptive change and new rules are being written. The song needs protecting as the keystone of it all."
In their joint explanatory memoranda, IMPF and IMPEL elaborated on the historical grievances they refuse to see repeated in the artificial intelligence sector:
"After years of addressing unequal revenue splits in streaming, the independent publishing community sees generative AI as an opportunity to establish a fairer framework, not another market in which outdated assumptions about the value of masters and compositions are carried forward."
Addressing the technical reality of how AI systems interact with intellectual property, the organizations noted that songs cannot be relegated to secondary status:
"Whilst AI generation can take place in various forms that must be treated on their merits, any AI licensing model that begins by treating the song ‘as a minor share of value’ while allocating ‘the overwhelming majority of revenues to master rights’ is simply unacceptable. Songs, lyrics, melodies, and compositions are central to the training, prompting, generation, output, and commercialisation of AI music products. Generative AI licensing must reflect the distinct and essential contribution of songwriters and publishers."
Finally, extending an olive branch to the tech sector while maintaining firm boundaries, the groups stated:
"The independent music publishing community is ready to engage constructively with AI companies, technology providers, and licensing partners, but emerging AI business models must not repeat or deepen the structural imbalances that have shaped music publishing in the digital market."
Future Outlook: Transparency, Administration, and the Path Forward
As generative AI tools transition from experimental prototypes to ubiquitous consumer software, the implementation of the IMPF/IMPEL framework will face severe operational hurdles. The roadmap ahead depends entirely on how effectively rightsholders, collective management organizations, and tech developers can solve complex administrative and legal puzzles.
The Imperative of Radically Transparent Accounting
One of the most profound demands outlined in the framework is absolute transparency. Historically, digital music licensing has been plagued by opaque accounting practices, complex data silos, and administrative deductions that obscure how royalties are calculated. In the AI era, this lack of transparency is entirely untenable.
The IMPF and IMPEL guidelines stipulate that rightsholders must have complete clarity regarding:
- Exactly which catalogues and works are being utilized by AI models.
- How specific algorithms attribute value and weight to individual compositions during the training and generation phases.
- The exact nature of any deductions, administrative fees, or legal cost allocations applied to incoming revenues.
- A clean separation of accounts between past infringement settlements, ongoing model training fees, and future product exploitation royalties.
Navigating the Bureaucracy of Collective Licensing
A significant operational challenge lies in the fact that music publishers are frequently one step removed from direct digital negotiations. Across much of the international repertoire, collecting societies—rather than individual publishers—negotiate, collect, and distribute blanket license revenues. Ensuring that these collecting societies are equipped to handle the granular data requirements of AI licensing, without siphoning exorbitant administrative overhead fees, will require a complete overhaul of back-end society infrastructure.
At the same time, for Anglo-American and independent catalogues where direct licensing is common, publishers must maintain a united front. If individual independent publishers are pitted against trillion-dollar technology firms in isolated negotiations, the risk of asymmetric bargaining power remains dangerously high. Collective frameworks like the one put forward by IMPF and IMPEL provide smaller publishers with the collective leverage necessary to demand equitable terms.
Conclusion: A Defining Moment for Musical Authorship
The release of the IMPF and IMPEL generative AI licensing framework marks a pivotal turning point in the modern music business. It serves as a stark warning to the technology sector: the era of treating musical compositions as disposable raw data or minor inputs in machine learning pipelines is over.
By demanding equal valuation, radical transparency, and a structural departure from the inequalities of the streaming age, independent music publishers are fighting to ensure that human songwriters remain fairly compensated in the dawn of artificial intelligence. Whether tech companies will meet these demands—or whether protracted legal and economic warfare will define the next decade of music copyright—will depend on how aggressively rightsholders enforce this new baseline. One thing is certain: the rules of the AI music economy are being written today, and independent publishers refuse to let someone else hold the pen.
