Music platform SoundCloud recently updated its terms of service, explicitly stating that "without your explicit consent, we will not use your content to train generative AI models designed to replicate or synthesize your voice, music, or likeness. Your consent must be clearly provided through an opt-in mechanism." This new policy, confirmed on May 11, 2025, by SoundCloud's Head of Communications, Marni Greenberg, to The Verge, seems to address user concerns about their uploaded content being used for AI training. It also promises transparency and a clear opt-out option if content is used for generative AI in the future.
Prior to this, SoundCloud revised its terms of service in February 2024 to allow uploaded music to be used for AI training without explicitly informing users. AI copyright activist Ed Newton-Rex discovered this change and sparked discussions about user data privacy and copyright issues. SoundCloud subsequently stated that it currently does not use artists' content to train AI models or build its own AI tools, and has blocked third-party scraping of data. It claimed that AI is only used internally for recommendations, fraud detection, and content classification functions, with artists retaining control over their works, and all AI usage adheres to existing licensing agreements. However, this statement did not completely rule out the possibility of using user data for more general AI model training.
Although the new terms limit the replication of individual voices, music, or likenesses without explicit consent, they remain unclear regarding how exactly user-uploaded content will be used for AI training. The article points out that generative AI models rarely directly mimic a single voice or style but still learn patterns from individual creators' works. In large datasets, the boundary between collective styles and personal traits becomes blurred, especially in music and sound fields, where influence and originality are often difficult to distinguish.
SoundCloud's stance is similar to OpenAI's concept of "studio style," which considers certain styles as shared characteristics of a studio rather than unique creations of specific artists, making them suitable for AI training. This view is controversial, particularly when a studio's signature style is often shaped by individual creators, such as Studio Ghibli and its animation master Hayao Miyazaki.
While SoundCloud's new terms increase user control over the use of their voices and likenesses for AI training, further clarification is needed regarding broader data usage and the definition of collective styles versus personal traits. The actual impact of its policies and how users can opt in or out of AI model training will be key areas of focus in the future.