SoundCloud has drawn criticism after a February 2024 update to its Terms of Service included language allowing user content to be used to “inform, train, [and] develop” artificial intelligence. The clause was flagged by Ed Newton-Rex, a former Stability AI executive and current CEO of Fairly Trained, who posted about the issue on social media.
The terms state that users “explicitly agree” their content may serve as input for AI technologies. However, a following line claims neither SoundCloud nor third parties can use content for AI training “without authorization from the applicable rightsholders.” This ambiguity has sparked concern among artists, particularly over the possibility of opt-out, rather than opt-in, mechanisms.
SoundCloud responded by asserting that it has never trained AI models on user uploads or allowed scraping for such purposes. The company said the updated terms were meant to clarify internal uses of AI, such as personalized recommendations, content moderation, and fraud detection.
Still, SoundCloud noted it would “keep our community informed” about future AI developments. Communications head Marni Greenberg stated that if the company ever moves to train generative AI models with user content, “clear opt-out mechanisms” would be introduced in advance.
Critics, including Newton-Rex, argue that an opt-out model is inadequate. He described it as unfair to artists, saying most would miss the chance to opt out, and called the clause a tactic to maximize training data. He announced his decision to delete his SoundCloud account, although it remained online at the time.
The situation raises broader questions about how digital platforms may handle artist content as AI integration expands across the music industry.
