YouTube's new setting lets creators opt their content into third-party AI model training
YouTube has launched a new setting that opts creators out of third party model training and enables them to choose whether they want all, some or no third-parties to train models using their content.
In September, YouTube announced it was working on several tools to help creators, on one hand, protect their likenesses and content from any unauthorized use while, on the other, also letting them manage whether and how their content should accessed by third parties looking to train AI models. In that context, YouTube stated that it would be able to share more about how it planned to prevent unauthorized scraping for model training purposes in a few months.
A recent announcement published in the YouTube Community tab of the service's help center delivers on that commitment; it also details a new setting in YouTube Studio enabling creators and rights-holders to control access to their content by third parties. Specifically, the new setting lets creators decide if third parties can access a video for model training purposes and, based on their decision, grant access to individual third-party companies, all of them, or none.
YouTube's permissions currently support granting individual access to 18 popular AI companies and an option to grant permission to all, which, in this case, means the 18 companies on the list plus any other interested third party. According to the announcement, for a video to be eligible for AI training, it must be allowed by the creator and the rights-holders of any content appearing in the video, including content detected by ContentID, which is YouTube's effort to mitigate copyright infringement within its platform.
The new setting is turned off by default, which is a sensible choice, given that it assumes most creators may not be interested in granting access to their content for model training purposes. Unfortunately, it seems unlikely that this setting will have any consequence in cases where the content has already been accessed. Moreover, the setting does not alter YouTube's Terms of Service, which explicitly reject unauthorized content scraping on the platform. It also does not affect agreements creators may have with Google regarding model training, and the company plans to continue using content to train AI models following these independent agreements.
