“Fair use” has become the weapon of choice for tech companies racing to train AI on copyrighted music, wielding the term as a battering ram while they deliberately sidestep the complex legal frameworks that protect artists and composers. This legal maneuvering is creating a dangerous precedent, opening the door to a future where master recordings can exist without an underlying composer, severing the chain of ownership and compensation that has defined music for a century.
Ilya Tolchenov, CEO of music-tech company Delphos, offered a stark warning about the industry's trajectory and a practical path forward that empowers creators rather than replaces them.
“The biggest nightmare scenario for musicians is a master recording with no underlying composition—something generated by tools like Suno—legally recognized and even copyrightable. But the reality is there's always a composer underneath it when the tool is trained on human compositions," said Tolchenov. "That’s when true creative replacement begins. The fact that these parallel issues are being pushed in different directions is what’s truly unethical."
While the rest of the industry remains locked in legal battles, Tolchenov is focused on tools that fit the existing music infrastructure. Delphos lets individual composers train their own models on just a handful of their own tracks, generate new music from them, and still claim composition copyright. The approach helps composers expand their catalogs and create new opportunities without replacing the human behind the work.
-
A practical alternative: "It sidesteps that whole legal question of, 'Is this fair use?' Because no, none of it's fair use," Tolchenov stated. "You can't just take my music and train on it. That should be what the conversation is. But this is a more practical way of approaching it: let's not just argue in court, let's just go protect our copyright."
There was always this idea with GarageBand that now everyone will be able to make music, but that never really happened. AI is similar. Yes, it lowers the barriers, but it's not making someone a musician who wasn't already going to be one.Ilya Tolchenov - CEO | DelphosThe philosophy directly counters the popular narrative that AI will "democratize" music creation, unleashing a flood of new artists. Tolchenov argued this is a fundamental misunderstanding of what it means to be a musician. The real barriers aren't technological; they are rooted in the identity, sacrifice, and emotional toll required to build a career in music.
-
The GarageBand fallacy: "There was always this idea with GarageBand that now everyone will be able to make music, but that never really happened," he explained. "AI is similar. Yes, it lowers the barriers, but it's not making someone a musician who wasn't already going to be one." Instead, he defined AI's true role not as a democratizer, but as a catalyst for professionals who are already on the path.
This leads to a crucial distinction about the nature of the tools themselves. According to Tolchenov, the AI solutions that will endure won't be consumer toys for generating disposable tracks, but professional instruments designed to solve tangible problems for working creators.
- Solving headaches, not making hits: "The best tools—and the ones that will last—in the AI space are going to be things that solve headaches, whether that's creative headaches or practical headaches," he said. "Delphos is going for the practical headaches, and we're aimed at professional composers. We focus on composition development, catalog expansion, and genre transfer."
Ultimately, the panic over AI-generated content flooding the market misses a larger, pre-existing reality. The challenge of discoverability—of fighting for attention in an infinitely crowded space—is not a new problem created by artificial intelligence. It's an acceleration of a crisis that has defined the streaming era for years. The real measure of AI in music won’t be how many new tracks it churns out, but how effectively it strengthens the hands of those already creating.