YouTube unveiled a significant update on Monday aimed at granting content creators greater authority over the use of their videos in training third-party artificial intelligence (AI) systems. This announcement follows the company’s earlier introduction of tools designed to safeguard creators against deepfakes that mimic their appearances and voices. The latest feature enables creators to control whether they permit AI companies to utilize their videos for training large language models (LLMs).
YouTube Empowers Creators to Control AI Access to Their Videos
As AI firms compete to gather data for training their models, many are seeking new sources of high-quality material. With the pool of publicly available data dwindling, these companies are exploring innovative methods to obtain substantial datasets that enhance the capabilities of their AI systems.
While some AI companies have opted for content partnerships, this approach is often costly. Alternatives such as synthetic data, generated by other AI models, present their own challenges, as this data can sometimes lack quality and hinder the progress of emerging models.
This situation has led many companies to pursue collaboration with content creators in their quest for valuable data. For example, Grok is trained using public posts from X (formerly Twitter), while Meta AI utilizes public content from Facebook and Instagram for its training.
YouTube has emerged as a key target for AI organizations due to its extensive repository of user-generated content. With the demand for data increasing for video generation models, the content on the platform is becoming increasingly important. However, YouTube has robustly protected creators by prohibiting unauthorized access to videos for crawling and scraping purposes.
According to a support document released by the platform, the new feature will give creators the ability to select whether they allow AI firms to access their videos for LLM training. In the coming days, YouTube plans to update its platform with a new setting in the Studio Settings under the “Third-party training” section.
This setting will allow creators to permit specific AI companies to access their videos. The list of companies currently includes notable names such as AI21 Labs, Adobe, Amazon, Anthropic, Apple, ByteDance, Cohere, IBM, Meta, Microsoft, Nvidia, OpenAI, Perplexity, Pika Labs, Runway, Stability AI, and xAI. Creators also have the option to grant access to all AI companies simultaneously.
YouTube clarifies that only those videos acknowledged by creators and relevant rights holders will be eligible for AI training. Their existing terms of service remain in effect, ensuring that AI firms cannot improperly scrape video content from the site.
While the update does not detail any financial compensation for creators whose videos are used by AI firms, YouTube has indicated that it will strive to foster new opportunities for collaboration between creators and third-party entities.
