For years, the creative sector has operated on a clear, if imperfect, bargain: when you use a song in a film, on a radio station, or for a commercial, you pay the person who wrote or performed it. Generative AI has broken that simple link, turning massive libraries of human artistry into the raw materials used to teach software how to mimic our collective culture.
We are currently seeing a move toward 'attribution'—the process of tracking exactly which training data influences a specific AI-generated result. Companies like Sureel are developing software that labels music files, allowing owners to dictate whether their work can be used for training and then attempting to measure their impact on the final output to justify potential licensing fees.
Solving the causality puzzle
Attributing value in AI is a significant technical hurdle because models don't 'copy' data in the traditional sense; they internalize patterns. Think of it less like a photocopier and more like a student who studies thousands of paintings to learn a style. To attribute value, engineers must determine which 'lessons' from that student’s study—which jazz tracks or folk songs—were most relevant to the final sketch they produced.
Some advocates argue for complex systems that reward artists every time their influence shows up in an output. Others, like SourceAudio’s Drew Silverstein, remain skeptical, noting that these systems are easily manipulated or simply too flawed to function reliably. Because these algorithms are built by people, they risk becoming opaque black boxes, hiding private deals behind a veneer of mathematical 'fairness.'
Ultimately, no algorithm can magically solve the problem of cultural compensation because 'fairness' is a human value, not a technical output. While we might look for perfect software to handle payments, the most sustainable path forward may simply be blunt instruments: public policy, industry-wide licensing agreements, or even taxing AI companies to redistribute wealth back into the cultural sectors they rely on. The technology might settle on a method for attribution, but the decision on whether it is equitable remains up to us.
Liked this one? The next lands at breakfast.
Every story in tomorrow's AI news, rebuilt in plain English — five minutes, sources linked, free forever.
By joining you agree to receive Article's daily newsletter — unsubscribe in one click. Privacy