Apple Intelligence Not Trained on YouTube Content, Says Apple

Apple has confirmed that its OpenELM AI model, trained on a dataset containing YouTube subtitles, is not used in any of its commercial AI features. The dataset was used solely for research purposes and is not integrated into Apple Intelligence, the company's AI system. OpenELM was released on the Hugging Face Hub to advance open-source language model development and is available through Apple's Machine Learning Research website. Apple has no plans to develop new versions of OpenELM. Apple Intelligence models are trained on licensed data, publicly available data collected by its web-crawler, and data selected to enhance specific features. The "YouTube Subtitles" dataset was used by other companies, including Anthropic and NVIDIA, for AI model training as part of a larger dataset called "The Pile."