The music industry has been clear about its stance on training musical AIs: licensing deals are necessary. But what should these licensing deals look like? A new white paper by Professor Daniel J. Gervais from Vanderbilt Law School offers some insights on the matter.
Published by Fair Trade Music International, a non-profit organization backed by various collecting societies and music creator groups, the paper's key recommendation is the creation of a new remuneration right.
“The best way for creators to ensure a continuous stream of revenue from the use of their copyrighted works by generative AI (GenAI) applications is to be compensated when the datasets containing their works are used to create new content. This should take the form of a license,” explains the executive summary.
The crucial point here is that this right would apply to the output generated by AI models, rather than the material used for training. However, Gervais suggests it could coexist with a licensing regime for the training process itself.
This oversight could fall to rightsholders – labels and publishers – but the paper suggests that the new remuneration right for AI-generated outputs should be vested in the original human creator, who could then assign this right to a Collective Management Organization (CMO), administrator, or other relevant parties.
This proposal is backed by collecting societies, who would likely support the idea, despite the challenges of fairly distributing royalties. “The credibility of a new right would depend in part on its ability to reach its intended beneficiaries,” writes Gervais. “CMOs could distribute funds based on metrics to be determined. Ideally, usage data would be available for this purpose. Many large language models (LLMs) can be programmed to identify the source material.”
He acknowledges that this proposal would be "fought by the AI industry," particularly if it decreases model efficiency and raises costs. However, he draws a comparison to the automotive industry’s resistance to catalytic converters decades ago.
“When a technology threatens the livelihoods of millions of human creators and potentially hinders the emergence of a new generation of creators, it doesn’t seem unreasonable to suggest a similar approach,” Gervais argues.
He also suggests the use of “credible proxies,” similar to those used for private copying and public lending, or even deploying a “second AI” as a compliance or detection agent to report on the output produced.
The full white paper is available for reading. Some parts require more legal knowledge, but if you're interested in these issues, it's an intriguing read. Now, we’ll have to wait and see how the major stakeholders in the music industry – especially the big labels – respond to these ideas
Fontes: Musically, Fair Trade Music
Comments