Apple, Nvidia, Anthropic, and Salesforce have been accused of using YouTube subtitles to train their AI models without creators' consent. Investigations revealed that the dataset, known as 'YouTube Subtitles,' included transcripts from over 170,000 videos across more than 48,000 channels. Apple has clarified that it used the dataset solely for its open-source OpenELM model, which is not connected to any consumer-facing AI features like Apple Intelligence. EleutherAI, the nonprofit behind the dataset, aimed to democratize AI research but did not inform content creators. Concerns have been raised about the ethical implications and potential financial harm to creators. Apple emphasized its commitment to ethical data sourcing, stating that its AI features are trained on licensed or publicly available data.