By Samantha Sawyer, General Manager, Licensing and Technology Solutions, MassiveMusic
A recent German court ruling found that ChatGPT violated copyright law by training on unlicensed song lyrics. This is merely the latest flashpoint in a much larger tension-filled dynamic: AI is advancing faster than the frameworks needed to ensure it is ethical, transparent and sustainable. And unfortunately, many models were trained on unlicensed, poorly tagged, or low-quality music data scraped from the internet. The result is a shaky foundation, legally risky for platforms and brands, and utterly opaque for the artists whose work fuels the system.
For brands and agencies integrating AI into everything from sonic branding to music selection and audio production, this moment is both a warning and an opportunity to create an equitable landscape moving forward.
The invisible fuel behind your AI tools
When most people think about AI and music, they imagine generative tools that can create a jingle or remix a track. But that’s only one use case. Music datasets are foundational to how AI learns structure, mood, genre recognition and emotional resonance.
This matters to advertisers because better-trained models deliver better creative alignment, faster workflows and safer outputs. For instance, when an AI tool understands the emotional cadence of music, it can help match the right sound to a brand message, suggest placements that resonate, or even assist in building campaigns that feel culturally fluent rather than algorithmically flat.
Built on borrowed work, running on borrowed time
Currently, there’s no global standard for AI training rights. Metadata gaps make it nearly impossible to validate ownership or track usage. Many models rely on datasets that were assembled without permission, compensation, or even basic documentation of what’s inside them.
This creates cascading problems. From creators whose work is being used to train systems that may one day compete with them; to brands and platforms that develop campaigns powered by an AI tool trained on unlicensed data, opening themselves to legal, financial and reputational risk.
The advertising industry has spent years building transparency into supply chains and data usage. We can’t afford to treat AI training data, particularly music datasets, as an exception.
What rights-first AI looks like
The solution isn’t a matter of slowing AI down; it’s about building it correctly from the start. That means music used for AI training must be licensed, opted-in and traceable. Rights need to be cleared across both master recordings and publishing. Metadata must be robust enough to ensure creators are correctly identified and compensated. And the entire process needs to be auditable.
Ethical frameworks don’t stifle innovation; they enable long-term, scalable innovation. Clean, rights-cleared datasets produce more reliable, higher-quality AI models. They reduce the risk of problematic outputs. They give brands confidence that the tools they’re using won’t result in a lawsuit down the line.
The infrastructure we’re missing (and how to build it)
Building ethical AI at scale requires infrastructure that doesn’t fully exist, yet. The music licensing ecosystem, while sophisticated, wasn’t designed for AI training use cases. New models are needed to determine how rights are structured, reported and compensated.
Models that have a deep expertise in navigating global licensing relationships across labels, publishers, and independent rights holders. Datasets enriched with detailed metadata, not just artist and title, but genre, mood, tempo, cultural context and usage permissions. Delivery systems that allow AI teams to ingest data in formats they can actually use. And governance frameworks that give creators control: the ability to opt in or out, and if they opt in, the assurance that they’ll be fairly paid.
Licensing the input side of AI training is becoming more streamlined. But compensating rights holders at the output stage remains a tougher challenge, largely because the technology needed to accurately identify which works contributed to a generated result is still evolving.
The brand case for getting this right
For marketers, this is about more than compliance. Ethical AI is all at once a consumer expectation and a competitive advantage. Brands that can transparently say their tools are built on licensed, responsibly-sourced data will differentiate themselves.
Stronger, more reliable AI also means better creative work. Tools trained on high-quality, diverse, well-tagged music datasets produce outputs that feel more human, more nuanced, more culturally resonant.
And perhaps most importantly, building ethical AI is an investment in a sustainable creative economy. Advertising has always depended on artists, musicians, writers, and creators. If we want that ecosystem to thrive in an AI-powered world, we need to make sure it’s structured so creators can actually make a living and where AI acts as an assistant to creativity and is not the usurper of it.
AI will shape the future of creativity, advertising, and media. But only if it’s built on a foundation that respects the people whose work makes it possible. Protecting artists and creators isn’t a barrier to innovation; it’s what makes ethical, long-term innovation possible in the first place.
