YouTube Allows Creators to Opt-In to AI Training

YouTube Empowers Creators with AI Training Data Control

YouTube announced on Monday a new initiative granting creators greater control over how their content is utilized by third parties for AI model training. This update allows both creators and copyright holders to specify which external AI companies are permitted to leverage their videos for model development.

New Settings in YouTube Studio

A dedicated setting has been integrated into YouTube Studio, the creator dashboard. This feature enables creators to opt-in and manage permissions for third-party AI companies.

Creators will find a list of 18 companies to choose from, granting them authorization to train their AI models on the creator’s video content.

List of Authorized Companies

• AI21 Labs
• Adobe
• Amazon
• Anthropic
• Apple
• ByteDance
• Cohere
• IBM
• Meta
• Microsoft
• Nvidia
• OpenAI
• Perplexity
• Pika Labs
• Runway
• Stability AI
• xAI

YouTube selected these companies due to their active development of generative AI models and their potential for beneficial partnerships with creators. Creators also have the option to allow “All third-party companies” access to their data.

Eligibility and Access

To utilize this feature, creators must have access to the YouTube Studio Content Manager with an administrator role. They can modify their third-party training settings at any time through their YouTube Channel settings.

Addressing Creator Concerns

The introduction of this feature responds to growing concerns from YouTube creators. Many have expressed dissatisfaction with companies, including Apple, Nvidia, and OpenAI, training AI models on their work without prior consent or financial compensation.

YouTube previously indicated plans to address these issues, and this new setting represents the first step in that process.

Google’s AI Training and Terms of Service

While this setting controls third-party access, Google will continue to train its own AI models using YouTube content, adhering to existing agreements with creators.

The new feature does not alter YouTube’s Terms of Service, which already prohibits unauthorized access to creator content, such as scraping.

Future Developments

YouTube envisions this as a preliminary step towards facilitating creator compensation for AI training. Future plans may involve allowing authorized companies to directly download videos for training purposes.

Default Settings and Retroactive Impact

The default setting for all creators will be to disallow third-party training, clearly communicating to companies that any prior training occurred without permission.

YouTube has not determined if the new setting will have any retroactive effect on existing AI model training. However, the company maintains that its Terms of Service already prohibit unauthorized access to creator content.

AI Detection Tools and Announcements

This announcement follows YouTube’s September unveiling of plans for creator controls regarding AI training, alongside new AI detection tools. These tools aim to protect creators from unauthorized use of their likenesses in other videos.

Creators will receive notifications about the new feature in YouTube Studio on both desktop and mobile devices over the coming days.

DeepMind’s Veo 2

In a separate announcement, Google’s AI research lab, DeepMind, introduced Veo 2, a new video-generating AI model designed to compete with OpenAI’s Sora.