Apple has come forward to deny a report that Apple Intelligence may have been unwittingly trained using data taken from YouTube without permission after it was found that data collected by EleutherAI included YouTube subtitles from thousands of YouTube videos.
In a report earlier this week, it was revealed that an AI training dataset generated by EleutherAI had been created using subtitles from 174,536 YouTube videos from over 48,000 channels. The data, taken without permission and in violation of YouTube’s rules, was then picked up by companies including Apple, which reportedly used it to train its OpenELM model.
