Apple Clarifies AI Training Data Sources, Denies Using YouTube Content

The company confirms OpenELM used YouTube-derived data but stresses it was solely for research, not for powering Apple Intelligence.

Apple asserts that its Apple Intelligence AI models are trained on licensed and publicly available data, not YouTube content.
The OpenELM model, which did use YouTube data, was developed for research purposes only and will not be updated.
Apple has offered websites the ability to opt out of their data being used to train its AI models.
The launch of Apple Intelligence has been postponed, with the beta now expected to release later this fall.
EleutherAI, the provider of the controversial dataset, used web scraping to compile data, raising ethical concerns.