Apple, along with several other tech giants, trained their artificial intelligence (AI) models using YouTube videos of famous creators like MKBHD without obtaining consent from them, it was revealed on Tuesday.
In what could be a bombshell in the tech industry, the website WIRED published an investigative piece claiming that companies including Apple, Anthropic, and Nvidia used AI training datasets that included videos from YouTube without creators’ prior consent.
The report suggests these companies used popular tech reviewers like Marques Brownlee (MKBHD), as well as content from other YouTube legends, such as MrBeast, PewDiePie, and various late-night talk show hosts to train their models.
MKBHD speaks out against Apple’s practices
In a tweet, MKBHD noted that “Apple has sourced data for their AI from several companies. One of them scraped tons of data/transcripts from YouTube videos, including mine.” The popular tech reviewer explained that Apple has found a technicality to avoid any legal action as “they’re not the ones scraping.” They did, however, emphasize that this issue is not going away anytime soon.
The dataset that Apple and other tech giants allegedly used included YouTube subtitles which contained transcripts from over 173,000 videos. The report noted that these videos were sourced from a staggering 48,000 or more YouTube channels.
What MKBHD explained in his tweet is that these videos were reportedly not downloaded by Apple itself. Rather, Apple used a non-profit organization called “EleutherAI.” This organization aims to help developers train AI models, and Apple allegedly used their services.
Nonetheless, the use of this dataset by Apple, as well as other companies, appears to clearly violate the terms of service of YouTube. According to these, harvesting of materials from the platform without permission of the creators is strictly prohibited.
MKBHD, who is being followed and viewed by millions of people around the world, expressed his concerns about this alleged unauthorized use of his content in a second tweet. There, he stated that he pays “a service (by the minute) for more accurate transcriptions of [his] own videos, which [he] then upload[s] to YouTube’s back-end. So companies that scrape transcripts are stealing *paid* work in more than one way.”
The revelation has already managed to spark a heated online debate about the strong ethical implications that AI training entails, including the rights of content creators.
Apple has not yet made any public comments on the matter.
Related: US Government Sues Apple for Manipulating Market