Generative music applications have achieved something that seemed unthinkable a few years ago: allowing anyone, with just two prompts, to can produce complete songs with vocals, arrangements and structures that can sound surprisingly real to most who hear them. This experience, which is presented as magical and accessible, has a much less visible side, linked to how these models have been trained and their legal implications. Many of these platforms have turned to large volumes of content available on the weboften copyrighted, to build their systems. The user enjoys the result, creates and shares, until a legal change, an agreement or a lawsuit transforms the tool and the experience is no longer the same.
Until just a few weeks ago, udio It was one of the services that best represented that promise of instant creativity. It had managed to attract both curious people and experienced musicians thanks to its simple system, the tools to extend, mix or remake songs and, above all, the possibility of downloading songs for use outside the platform. There was nothing to suggest that this model was about to change. The first indication came when the company began to talk about a “transition phase” linked to new agreements with record companies. It did not yet detail what was going to happen, but it made it clear that the platform was entering a different stage.
The day the download button disappeared. Confirmation came when Udio announced thatas part of its transition, audio, video and stem downloads would be disabled for several months. It was a feature that many considered essential, but now they could only play their creations in Udio and share them using links from udio.com. In exchange, the company reported an increase in credits and more generation capacity, although that did not compensate for the feeling of loss. The message was clear: the songs still existed, but they no longer left the walled garden.
Warner and Universal chose a different path than the judicial confrontation: turning Udio and Suno into partners rather than adversaries. Universal signed agreements for the next version of Udio to be based on licensed music and offer artists new avenues of income, while Warner did the same with Suno and also sold the Songkick platform to incorporate it into that new ecosystem. Record companies went from denouncing to collaborating, with a clear condition: at least in the case of Warner and Udio, artists and composers would have the possibility of deciding whether their voice, their image or their style could be part of the creations generated by AI.
From defendants to partners. Once the content is within the legal space, what is relevant is not only that agreements have been signed, but how the industry’s priorities have changed. A year ago the goal was to put AI platforms on the bench for using protected music to train their models. Today, a growing part of the sector has understood that it may be more profitable to integrate them than to stop them. The move does not eliminate legal conflicts, but it opens the door to a model in which record labels oversee, license and participate in revenue, rather than reacting only through lawsuits. It is a change of focus that signals where the music business is moving.
What nobody sees: scraping as the foundation of musical AI. For years, the actual functioning of many generative music models was far from transparent. Some startups, like Suno, admitted to having trained their systems with “virtually every quality music file available on the web,” trusting that such use would be protected by the fair use. However, when record companies began to examine that process, the conflict ceased to be technical and became legal.
Images | Universal Music | udio | Unsplash
In Xataka | AI has become the best example that if you don’t pay for the product, you are the product
GIPHY App Key not set. Please check settings