YouTube announced Monday that it will give creators more options for how their content can be used by third parties to train AI models. Starting today, creators and rights holders will be able to report to YouTube if they allow certain third-party AI companies to train models on their content.
Creators will be able to opt in to this new feature if they want through YouTube Studio, a new setting in their Creator Dashboard. Here you'll see a list of 18 companies that you can choose to have permission to train with your videos.
Companies included in the initial list include AI21 Labs, Adobe, Amazon, Anthropic, Apple, ByteDance, Cohere, IBM, Meta, Microsoft, Nvidia, OpenAI, Perplexity, Pika Labs, Runway, Stability AI, and xAI. YouTube notes that these companies were chosen because they are building generative AI models and are likely smart choices for partnering with creators. However, creators can also select a setting called “All third-party companies.” This means allowing third parties to train your data, even if they are not on the list.
The company also said that eligible creators are those with access to YouTube Studio Content Manager, which has an admin role. You can also view or change your third-party training settings at any time in your YouTube channel settings.
With the rise of AI technology, especially AI videos like OpenAI's Sora, YouTube creators are increasingly worried about how companies like Apple, Nvidia, Anthropic, OpenAI, and even Google itself can use their material without consent or compensation. I complained that I was training an AI model. YouTube announced this fall that it would address this issue in the near future.
But while the addition of this setting controls third-party access, the company told TechCrunch that Google will continue to train its own AI models on some YouTube content, per existing agreements with creators. Ta. Additionally, these new settings do not change YouTube's Terms of Service, which prohibit third parties from accessing creator content through scraping or other unauthorized methods.
Rather, YouTube sees this feature as a step toward making it easier for creators who want to help companies train AI on their videos, and perhaps as a way to get paid for that training. In the future, YouTube may take the next step in this process by granting access to companies whose creators have granted them access to direct downloads of their videos.
With the introduction of this feature, the default settings for all creators will no longer allow third-party training on videos, making it more clear for companies that have already trained that they have trained against the creator's wishes. It will be.
YouTube could not say whether the new settings may have any retroactive impact on previously trained third-party AI models. But the company says its terms of service indicate that third parties cannot access creator content without permission.
The company first revealed plans to offer Creator Control for AI training in September, with the goal of preventing creators, artists, musicians, actors, and athletes from having their likenesses captured, including their faces and voices. It also announced a new AI detection tool. Copied and used in other videos. The detection technology expands on YouTube's existing Content ID system, which previously focused only on copyrighted material, the company explained at the time.
Creators around the world will be notified of new features via banner notifications in YouTube Studio on desktop and mobile in the coming days.
Separately, Google's AI research institute DeepMind on Monday announced a new video generation AI model called Veo 2 that aims to rival OpenAI's Sora.