AI and music copyright: A field guide for artists and developers

This report is part of our Season 3 research sprint on creative AI for the music industry.

Large-scale image generators (e.g. Stable Diffusion, Midjourney) and large language models (e.g. GPT-3) have been trained on vast swaths of allegedly copyrighted material. These tools have the potential to create outputs that directly compete in the marketplace with non-AI powered creation — and lawsuits have been picking up accordingly .

The music industry is watching what has happened with image generators closely, and voicing similar concerns around legal topics of attribution, compensation, and opt-in/opt-out functionality for rights holders.

Amidst the highly ambiguous legal landscape for creative AI, we wanted to provide a more accessible, foundational state-of-play resource for artists and developers who are eager to experiment, but unsure of their rights and risks.

As a community, we reviewed the Terms of Service for a diverse group of 11 creative AI tools — ranging from the most well-known image generation tools like Midjourney and DALL-E, to emergent music AI startups like Boomy and AIVA. We also read various scholarly writing in the developing area of AI and the law, including articles that considered detailed jurisdictional approaches, issues, and case law, to provide important historical context.

Our analysis seeks to fill gaps in existing research by providing a primary analysis of existing AI platforms’ policies , with a specific focus on those that are part of the recording artists’ emerging tool set . We aim to educate artists on the state of the law and industry norms around these tools, including the extent of their rights, ownership, and protection, and the implications when different stakeholders generate revenue.

You can view the full report below (recommended to view on desktop), and catch up on the rest of our creative AI research on our STREAM landing page . Thank you so much for your support!