Ethan Klein is conducting a class action lawsuit against major tech giants for scraping millions of YouTube videos.
AI cannot learn from nothing, it has to be trained. So many major companies decided to use YouTube videos to train their AI models without compensating any of the creators. Ethan Klein has now conducted a class action lawsuit against them for illegally aquiring millions of videos.
"AI Is Stealing People's Content"
On April 15, Ethan Klein appears alongside his lawyer AttorneyTom in an episode of Ethan's H3 Podcast titled "AI Companies Stole Our Content. I'm Taking Them to Court."
In it, the influencers announce a new lawsuit against major companies Nvidia, Meta, Bytedance, Snap, Amazon, OpenAI, Apple, and maybe more to come, for illegally scraping YouTube videos for their AI models.
This lawsuit was initially triggered by a 404media article that leaked internal conversations of Nvidia indicating that the company used YouTube videos to train their AI with executive approval. According to Ethan, the companies are able to scrape 80 years worth of content a day.
They allegedly take the YouTube videos to mass train AI models for them to make videos themselves without compensating the original creators, even though once released, the companies will make "trillions of dollars" off of these models. Ethan says the companies steal people's content off YouTube. They use YouTubers to train AI models to replace the YouTubers.
This lawsuit is a class action lawsuit on behalf of all the creators, whose content was scraped. Ethan and his attorney Tom claim they have the datasets that show which videos were used in the training. H3 alone supposedly has multiple hundreds of videos in these datasets.
The lawsuit has been in the works for months now. Ethan asserts that this lawsuit is not about the money but about holding these big companies accountable for their wrongdoings.
This whole phenomenon is relatively new, which is why there are no explicit rules and laws regarding the scraping of YouTube videos. The companies, as stated by Ethan and Tom, seem to hold a systemic belief that since YouTube videos are free to watch, they can do whatever they want with them because you can not steal what is free. They also have allegedly referred to fair use as the model's output is different enough from the originals as to not count as a copy.
But they miss the point. The videos are free to watch on YouTube, as YouTube compensates creators for every click, but copyright still remains. Creators have copyright on their video, even if it is available on YouTube. The AI models using the videos to train do not "watch" them. The infringement lies in the way in which they aquire the videodata. They are, as Tom says, "breaking the lock of YouTube", intentionally and specifically bypassing YouTube's technological protection measures to extract the data.
So the problem is not the fact that the AI models watch and copy the videos or the output they create, but the way in which they get the data to create the system.
This is a nationwide class action lawsuit for violations of the DMCA's anti-circumvention provisions (...) arising from the Defendant unlawfully circumventing technological protection measures to access and scrape millions of copyrighted videos from the online video viewing platform YouTube in order to feed, train, improve and commercialize Defendant's large scale generative text-to-video artificial intelligence model.
the lawsuit against Apple states, adding:
Rather than seek permission or pay a fair price for the audiovisual content hosted on YouTube, [Apple] harvested content creators' protected and copyrighted videos for commercial use and at scale without consent or compensation to the content creators.
The YouTubers are seeking statutory damages, injunctive relief, and restitution.
[Apple's actions were] not only unlawful, but an unconscionable attack on the community of content creators whose content is used to fuel the multi-trillion-dollar generative AI industry without any compensation.
There is a lot at stake here. The comanies are at risk of loosing their product because they can not make it without hundeds of millions of these videos, and if unchecked they could use the hard work of creative people to make soulless AI content replacing passionate projects without even reimbursing them.
But the lawsuit seems to be strong with the datasets available to Ethan and his lawyer and even YouTube's CEO stating in previous interviews that scraping YouTube videos is absolutely against the site's terms of service. It will stay exciting to see what this lawsuit will change in the use of people's content and creations in AI training, and if the companies will be held accountable.
