A thorough investigation by 404 Media revealed that Runway’s Gen-3 artificial intelligence model may have scraped thousands of YouTube videos without permission, including from official Nintendo channels.
The news outlet obtained a spreadsheet listing channels that were targeted- and Nintendo was not the only big name to be on there, with Disney, Rockstar Games, Netflix, and Sony also being affected. Part of it is pictured below, highlighted by Nintendo Life:
Runway has seen financial support from both Alphabet- Google’s parent company- and Nvidia, and its Gen-3 model was met with considerable praise during its launch.
While 404 Media was unable to confirm if every single video listed in the spreadsheet was scraped, it was able to prove that Runway did intend to use the list of videos to train Gen-3. A statement from a former Runway employee, whose identity was withheld for their safety, all but confirmed it:
“The channels in that spreadsheet were a company-wide effort to find good quality videos to build the model with. This was then used as input to a massive web crawler which downloaded all the videos from all those channels, using proxies to avoid getting blocked by Google.”
When asked to comment by 404, Google, both the owner of YouTube and involved in supporting Runway, said that their statement earlier this year, where they said that AI being trained off of YouTube videos was in violation of the platform’s rules, still stands.
Even if Nintendo was not directly affected, or chose not to pursue Runway, the other victims may not be as forgiving. And what role Google will play in all of this is yet to be determined- as well as the future of Gen-3 and all other AI generators like it.
And what exactly would scraping the youtube videos get them?
My understanding is that “feeding” these videos to the AI allows it to learn and improve upon its generative abilities. Gen-3 seems to focus on generating videos through artificial intelligence, so scraping other non-AI videos allows it to generate videos that are higher in quality and may even appear to be human-made at first glance. However, someone better informed than me can correct me or add to what I’ve said below.
It’s concerning to hear that a generative AI model may have scraped Nintendo’s YouTube videos without permission. This highlights the ongoing ethical debates surrounding AI training data and copyright. It’s a reminder that developers, like those behind DeepSeek (more information at DeepSeekDeutsch.io), need to be transparent and responsible in their data acquisition practices to avoid potential legal and ethical issues.
The way I see it, if you don’t want your info to be used then don’t post it. This goes for companies and individuals alike.
Maybe this will result in one of these big companies suing over IP infringement and we’ll finally get some laws and regulations on AI.
That should probably fit the bill over AI if that law is written. Beside’s I have nothing against AI at all.
Naturally not; after all, you are one.
Do you mean “scrapped”? Haha.
If true, this raises serious questions about how generative AI models are trained. Even advanced systems like DeepSeek (https://deepseekturkce.com/) need to respect content rights. Nintendo’s approach to IP is strict for a reason, and scraping without permission crosses the line. AI progress shouldn’t ignore creative ownership.