As artificial intelligence rewrites the rules of the music industry, an a16z-backed startup is betting on two fronts at once. Its focus: recovering money artists never collect, and licensing catalogs to train the machines that will make the music of the future.
There's a paradox in music that few outside the trade are aware of: every year, enormous sums of royalties are generated that never reach their rightful owners. It isn't out of bad faith, but something more mundane—poorly registered works, incomplete data, and records that simply don't exist. That invisible money is scattered across collecting societies worldwide. This is the first territory
Unisound aimed to conquer. The music-tech company, headquartered in Miami, was founded by Gregorio Torres and Andrés Ginebra.
The pitch is straightforward: put artificial intelligence agents to work across catalogs to detect discrepancies, unclaimed royalties, and missing registrations, and then help publishers collect what they're owed on an international scale. It is, in essence, an automated and permanent audit of copyright. On that front, Unisound already works with publishers such as 1844 Publishing, Black 17 Publishing, and Koze Music Publishing. The company previously conducted a pilot project with Sony Music Publishing Latin focused on identifying potentially unclaimed royalties. The pilot identified millions of dollars in potential royalties, underscoring the value of precision in rights and royalty analysis.
But the most revealing clue about where the industry is heading lies not in royalty recovery, but in the company's second line of business. The explosion of AI models capable of generating music and audio created an urgent and delicate need: these systems must be trained on real catalogs, and doing so legally requires securing rights, attributing content, and compensating the people who created it. The tech companies themselves began looking for intermediaries capable of tracking millions of songs and managing those rights. Unisound positioned itself squarely in that bottleneck.
The result is a company with an unusual dual model: it recovers royalties on one side and licenses music for AI training on the other. On that second pillar, there are already signed agreements with LANDR, and Serato, as well as ongoing confidential conversations with global technology platforms. The LANDR case illustrates what the future of compensation might look like in the AI era: through the data intermediary ERISV, Unisound takes part in LANDR's Fair Trade AI program, an initiative that allows content providers to earn recurring income from the use of their material to train AI-assisted music tools. LANDR itself describes it as the music industry's first mature opt-in attribution model: rather than machines being trained on others' music without permission, creators choose to participate and get paid for it.
That shift—from forced opt-out to compensated opt-in—is perhaps the most important cultural battle being waged in the music business today, and it explains why a fund like Andreessen Horowitz (a16z) decided to invest in a company of barely three people. For its founders, that bet carried a meaning that goes beyond the financial: it validated the idea that the music industry now competes in the same league as tech, finance, or biotech startups. Put another way, music has stopped being treated as a low-margin cultural sector and entered the radar of elite venture capital.
What's at stake is significant. As artificial intelligence produces more of the music we listen to, the central question shifts: it is no longer whether machines will learn from human catalogs, but under what rules and with what compensation they will do so. To address this, companies like
Unisound are building, in real time, the infrastructure for those rules—determining who gets paid, how content is attributed, and how human talent retains its value in an increasingly automated ecosystem. At this frontier—between the money never collected and the money just starting to be generated—the economic future of music is being decided.
