A recent report by Proof News and Wired reveals that major tech players like Apple, Salesforce, and Anthropic have trained their AI models using tens of thousands of YouTube videos without creators' consent. The data was sourced from "the Pile," a collection by EleutherAI, which includes YouTube captions from popular channels like MrBeast and PewDiePie. This scraping of content has raised concerns about intellectual property rights and potential violations of YouTube's terms of service. Despite being used in the past, the specifics of this scraping have now come to light, prompting backlash from content creators like David Pakman and Julia Walsh. While some companies argue that there was no violation, the issue of unauthorized scraping of YouTube content remains unresolved. As AI-generated content becomes more prevalent, the need for ethical data collection practices is emphasized, highlighting the complex challenges surrounding the use of online content for AI training.