On Monday, YouTube unveiled a significant update aimed at granting content creators increased authority over the use of their videos for third-party artificial intelligence (AI) training. This announcement follows the platform’s recent introduction of tools designed to safeguard creators from deepfakes that mimic their appearance and voice. With the new feature, creators will have the ability to permit or deny access to their videos for training large language models (LLMs). They can selectively grant permission to certain AI firms while restricting others from utilizing their content.
YouTube Empowers Creators on AI Firm Access to Their Videos
As companies compete to gather fresh data for training AI models, the demand for high-quality sources has intensified. With publicly accessible data becoming limited, many AI firms are searching for alternative means to acquire substantial datasets that enhance the functionality of their models.
While some AI organizations have opted for content partnerships to access this necessary data, such arrangements are often costly. Another alternative involves the use of synthetic data generated by other AI systems, though this method carries the potential for producing lower-quality datasets, which could hinder the development of newer models.
In light of this, organizations are increasingly seeking collaborations with content creators to uncover new sources of high-quality data for AI training. For example, Grok is trained using public posts from X (formerly Twitter), while Meta AI relies on public posts from Facebook and Instagram.
YouTube is also becoming an appealing platform for AI companies due to its vast repository of human-generated content. The rise of video generation models has further amplified the value of this data. However, YouTube has thus far prohibited unauthorized crawling and scraping of videos to ensure creator protection.
According to a support document shared by the company, this new option will empower content creators to decide on permitting AI firms to use their videos for LLM training. YouTube intends to roll out this feature within days, adding a new section in Studio Settings labeled “Third-party training.”
Creators will then have the choice to allow specific AI companies to scrape their videos from a predefined list that includes AI21 Labs, Adobe, Amazon, Anthropic, Apple, ByteDance, Cohere, IBM, Meta, Microsoft, Nvidia, OpenAI, Perplexity, Pika Labs, Runway, Stability AI, and xAI. Additionally, creators can elect to grant broad access to all AI firms by selecting the corresponding option.
YouTube emphasizes that only those videos sanctioned by creators and relevant rights holders will qualify for AI training. The platform’s standard terms of service will still apply, ensuring that AI firms cannot improperly scrape videos from the site.
Notably, this new feature does not reference any form of compensation that AI firms might provide to creators for the use of their videos. Nevertheless, YouTube reiterated its commitment to fostering collaborative opportunities between creators and third-party companies moving forward.
