"Four datasets containing millions of music tracks have been identified as being used for AI model training. The datasets include works from major artists and are being used by companies like Google and Stability AI. The collection methods used to compile these datasets may violate platform terms of service."