Proof, in collaboration with Wired, has published an investigative article shedding light on the case of several tech giants such as Apple, Anthropic, NVIDIA, and Salesforce allegedly using data from YouTube subtitles to train their AI models without permission. This unauthorized data scraping has raised concerns, with OpenAI previously facing similar accusations.
According to Proof, over 173,536 YouTube clips from 48,000 channels, including popular YouTubers like MKBHD, Mr. Beast, and Pewdiepie, as well as news channels such as BBC and The New York Times, had their subtitle data harvested for AI training purposes.
Despite this revelation, MKBHD noted that companies like Apple typically purchase data from other companies, raising the possibility that these tech giants might have acquired YouTube data without proper authorization to resell (and other accused companies potentially engaging in similar practices?)
Source: Proof, MKBHD
TLDR: Apple and other tech giants are using YouTube subtitle data to train their AI models without permission, sparking concerns about data privacy and unauthorized use.
Leave a Comment