The music industry's infrastructure paradox. When the money moves faster than the data.

Money will always move faster then data, if copyrights royalties is stuck in a fax machine.
The convergence of AI-generated content and stablecoin payments presents a short window to resolve a long-standing issue or make it permanently worse.

On February 24, 2026, Google announced the acquisition of ProducerAI (formerly Riffusion) and integrated its team into Google Labs and Google DeepMind. It also merged its conversational music creation platform with Lyria 3, the company's most advanced generative music model.
That same week, Meta was reported to be circulating proposals for a stablecoin-based payment system, aiming for deployment in the second half of 2026 across its 3.58 billion daily active users.

Each development matters on its own. When combined, they highlight a specific, urgent problem that the music industry hasn't solved and is about to undergo a stress test on a scale it has never experienced before.

What Google actually built

Lyria 3 is more than just a small update. It creates full vocal tracks with automatically generated lyrics, supports multiple languages, responds to text and image prompts, and embeds a SynthID watermark into every output—a signal that can survive MP3 compression, pitch shifting, and speed changes, and that Gemini can detect in uploaded audio. The Gemini app integration lets any user generate a track from a photo, refine it through conversation, get custom artwork from the Nano Banana model, and share it on YouTube Shorts—all within one workflow.

ProducerAI's contribution to this stack is the conversational layer. Co-founder Seth Forsgren built a platform where music creation functions as it naturally does, through iterative adjustment, revision, and gradual refinement, rather than via single-shot prompts. Co-founder Hayk Martiros, now at Google DeepMind, combines engineering expertise with practical musicianship. The Chainsmokers specifically highlighted that musician-engineer blend as the platform's key feature when they invested in the seed round.

Understanding the technical origins of the entire project is important. Riffusion launched in December 2022 as an open-source experiment: it used Stable Diffusion v1.5, fine-tuned it on mel-spectrograms—visual representations of audio that display frequency over time as 2D images—and then generated new spectrograms from text prompts, converting them back to audio via an inverse Fourier transform. The idea was innovative and creative. The legal risk was immediate and fundamental: Stable Diffusion's training data was gathered from web scraping, meaning Riffusion's sound vocabulary was based on audio it did not have the rights to use.

The ProducerAI rebrand in July 2025 publicly indicated that the team understood this. The Google acquisition is the main solution: Lyria 3 is trained on data licensed from Universal Music Group and sourced from the YouTube catalog under bilateral agreements, placing it in a much different legal position from competitors like Suno and Udio, both of which are facing ongoing mass infringement litigation.

What Google did not build

Here is the gap that the acquisition cannot fill, and that no bilateral licensing agreement between a platform and a major label can close.

Google's attribution system, which traces an AI-generated output back to the training data that influenced it most, operates within Google's ecosystem. It recognizes the catalog licensed by Universal and content registered in YouTube's Content ID. However, it does not know about the independent producer whose drum programming helped define a sub-genre, the session pianist whose harmonic voicings appear throughout the model's learned vocabulary, or the unsigned songwriter whose melodic patterns the model absorbed from a SoundCloud upload.

Google argued to UK and US regulators as recently as January 2026 that training on open-web data should not require compensation if the output is "non-expressive," meaning it does not directly reproduce a melody or name a specific artist. Whether that argument holds up under legal scrutiny is still being debated. In practice, what this means is clear: the compensation model being developed by major platforms applies to those with contractual power and considers everyone else outside the payment system.

The music industry faces a layered payout issue embedded in the structure of AI-generated content, even before stablecoin settlement occurs.

What stablecoins do to an already unequal system

Meta's 2026 stablecoin approach differs significantly from the 2019 Libra attempt. Instead of building a proprietary monetary network, which faced swift resistance from central banks over sovereignty concerns and eventually collapsed under political pressure, the current plan connects to existing regulated systems through licensed third-party intermediaries. The GENIUS Act, signed in July 2025, created a federal framework for payment stablecoins in the US, including reserve requirements, Bank Secrecy Act obligations, and enforcement powers. Stripe's Bridge entity received conditional OCC approval in February 2026 to operate as a national trust bank. The regulatory framework that Libra lacked is now in place.

This means Meta's distribution goal isn't blocked by the absence of a compliant stablecoin framework. It's operational planning, not regulatory approval, that sets the current moment apart from deployment.

In music, the implications aren't mainly about consumer behavior but about transaction infrastructure. Stablecoin settlement at scale enables three things that traditional systems can't: near-real-time cross-border creator payouts without high fixed-cost conversion fees, programmable micro-licensing where payments automatically trigger upon usage detection, and conditional payment logic, such as escrow until disputes are resolved, territorial gating, or split payments among contributors, which batch accounting systems cannot handle.

These capabilities are truly valuable. They also enhance the existing features of the underlying rights data. A properly attributed catalog with machine-readable ownership splits is fully automated; each usage event immediately triggers a correctly routed payment, at the correct rate, to the appropriate parties, in the correct territory. An incorrectly attributed catalog leads to repeated misallocations: the same errors that currently result in slow, incorrect payments will instead generate fast, inaccurate payments on a scale that makes manual correction impossible.
The music industry currently receives about 100,000 new track submissions daily for streaming platforms. The Mechanical Licensing Collective exists specifically to manage the large volume of unmatched mechanical royalties that the current system cannot process, recognizing that data mismatch at this scale is a fundamental aspect, not an anomaly. Apple Music reported demonetizing two billion fraudulent streams in 2025. These issues are common; they form the baseline condition into which stablecoin velocity is about to be introduced.

The registration layer that is missing

The problem has a clear structure and a known solution plan.

Each musical work requires an ISWC, and each sound recording requires an ISRC. Every contributor, whether a composer, lyricist, performer, producer, or arranger, needs a persistent identity record that accurately resolves across aliases, pseudonyms, transliterations, and multi-script name variants. Ownership splits should be encoded as structured data fields, not as PDF attachments or narrative clauses in contracts. Each rights set should be scoped territorially and by rights type at registration, rather than reconstructed on a case-by-case basis when a use occurs.

This is not a speculative architecture. The UK government's industry metadata agreement, DDEX's Recording Information Notification standard, and the Music Modernization Act's blanket license framework all identify the same gap from different perspectives: the data needed to route payments accurately must be captured and validated at the moment of creation and registration because reconstructing it later becomes increasingly costly as more platforms, use types, and transaction volumes enter the system.

The challenge is that no neutral, interoperable registry currently performs this function at the standard that automated settlement requires. Instead, what exists is a collection of bilateral agreements, platform-specific matching systems, and collecting society databases built for a batch-accounting world that are now being asked to interface with settlement infrastructure designed for real-time execution.

SynthID functions as Google's watermark, while Content ID acts as Google's matching layer. The attribution engine that traces Lyria 3 outputs to training contributors is an internal Google tool. Each system is technically capable and highly sophisticated in operation, but each is compatible only with the parties Google has contracted with. An independent creator with a fully documented rights record in a neutral registry has no automatic access to Google's attribution engine, no guaranteed visibility in Meta's payment routing, and no way to participate in the AI training compensation frameworks being negotiated between platforms and major labels unless a neutral, platform-agnostic infrastructure layer exists to connect them.

What correct registration actually means in this environment

The window for establishing that infrastructure layer is closing. As AI attribution systems improve and become more focused on existing data relationships, creators whose rights are not documented in a machine-readable, standards-compliant, identity-verified format will find themselves outside the payment system. This is not because any platform intentionally excludes them, but because systems optimized for speed and automation rely on the data that is available.

Proper registration fully requires five elements that are simple individually but rarely seen together: standard identifier anchoring at both the work and recording levels; meaningful contributor identity resolution that persists across name variations; machine-readable ownership splits encoded as executable data rather than text; territorial and rights scope; and cryptographic provenance that creates an auditable chain of titles from creation through every transfer or licensing event.

A catalog with those properties is inherently designed to be machine-routable. When Meta's payment infrastructure operates at stablecoin speeds across 3.58 billion daily active users, when Google's Gemini agents make licensing decisions automatically within commercial content workflows, and when AI-generated music traces its training DNA back to contributing human creators, a properly registered catalog automatically participates in all these processes.

Everything else requires manual intervention. With the current transaction volumes entering the system, relying on manual intervention is not a sustainable operating model.

The structural opportunity

The economic logic behind this is straightforward. The surplus in the music rights market is flowing to those who handle the settlement, provenance, and compliance infrastructure, not because these operators charge rent, but because the tasks they perform are essential for the revenue of every other participant in an automated settlement system.
For independent creators, the main takeaway is clear:
registration quality now significantly impacts financial matters, not just administrative ones.
The infrastructure is being built around existing data.
Creators who have established documented, auditable, machine-readable provenance before stablecoin settlement is fully implemented are well-positioned to participate in the automated economy being developed. Those who haven't set this up are likely to be reconciled manually later, at their own expense, against systems that weren't designed to wait for them.

The money is already flowing. The question is whether the copyright infrastructure is prepared to handle it properly.