Sources: OpenAI transcribed 1M+ hours of YouTube videos through Whisper and used the text to train GPT-4; Google also transcribed YouTube videos to harvest text (New York Times)

New York Times:
Sources: OpenAI transcribed 1M+ hours of YouTube videos through Whisper and used the text to train GPT-4; Google also transcribed YouTube videos to harvest text  —  OpenAI, Google and Meta ignored corporate policies, altered their own rules and discussed skirting copyright law …