YouTube is adding a new layer of control for creators whose videos could be used to train AI models. The platform is rolling out settings in YouTube Studio that let creators decide which AI companies, if any, can receive permission to use their content for model training.
The change gives creators a clearer way to manage AI training access from inside the tools they already use to run their channels. It also draws a sharper line between permitted use and unauthorized content scraping, which remains banned under YouTube's existing terms of service.
What the new YouTube AI training settings do
The new controls appear in YouTube Studio under Third-party training. From there, creators can choose which AI companies are allowed to train models using the creator's content.
According to Techcrunch, the current list includes AI21 Labs, Adobe, Amazon, Anthropic, Apple, ByteDance, Cohere, IBM, Meta, Microsoft, Nvidia, OpenAI, Perplexity, Pika Labs, Runway, Stability AI, and xAI.
The setting is not a blanket transfer of content rights to every AI company. Instead, it gives creators a company-by-company permission mechanism. A creator can allow some companies while withholding permission from others, and YouTube says creators can modify or withdraw those permissions at any time.
That matters because AI training has become a central issue for online creators. Videos can contain performance, instruction, commentary, music, visual work, and other creative material. A single channel may represent years of production, and creators increasingly want a say in whether that work becomes part of AI model development.
Approval is not only up to the uploader
YouTube's system does not rely only on the decision of the person managing the channel. For an AI company to train on a video, several conditions must be met.
- The creator must approve the training access.
- Any rights holders identified through Content ID must also approve it.
- The video must be public.
- The video must follow YouTube's guidelines.
This means a creator's approval alone may not be enough when Content ID identifies other rights holders. That is an important limit for videos involving material that may be claimed or controlled by someone else.
The public-video requirement also narrows the scope. The controls are not described as covering every possible video state across YouTube. Based on YouTube's conditions, training access depends on public availability and compliance with platform rules.
There is also a timing detail for developers and companies that rely on YouTube's public YouTube Data API. YouTube notes that when creators change their settings, updates can take up to seven days to appear in the public YouTube Data API.
What YouTube does not promise
The new controls create a permission path, but YouTube is also clear about the limits of its role. The platform says it will not control how AI companies ultimately use creator content after permission is granted.
That distinction is important. A creator may give a company permission to train on eligible videos, but YouTube cannot guarantee that the company will actually use the videos it has been allowed to access.
In practical terms, the setting is about authorization. It is not a full oversight system for what happens inside an AI company's training process. Creators gain a direct way to express permission, while YouTube avoids promising control over downstream use by third parties.
The feature also works alongside the platform's existing rules. YouTube's terms of service still ban unauthorized content scraping. The source article notes that this has been a point of tension before, including prior issues between YouTube and OpenAI over potential unauthorized data collection.
Why this matters for creators and AI companies
YouTube describes the new controls as an "important first step" to help creators gain value from their content in an AI-driven world. The phrase signals that the company sees these settings as an early move, not the final form of creator-AI collaboration.
For creators, the immediate value is control. Instead of having only broad platform rules in the background, they get a visible setting tied to specific AI companies. That can make decisions about AI training more explicit and easier to revise over time.
For AI companies, the controls create a clearer permission channel for YouTube videos that meet the stated conditions. Access still depends on creator approval, Content ID rights-holder approval where relevant, public video status, and YouTube guideline compliance.
The feature may also make consent easier to document through YouTube's systems. However, because updates can take up to seven days to appear in the public YouTube Data API, permission changes may not be reflected instantly in all places where outside systems check that information.
YouTube is also expanding its own AI work
The creator controls are arriving as Google continues to build AI features for YouTube itself. The source article says the company plans to launch Veo 2, a new video model, next year to help create short films and other content.
That places YouTube in two roles at once. It is giving creators a way to manage third-party AI training permissions, while Google is also expanding AI capabilities on the platform.
The key takeaway is that YouTube is trying to formalize how creator content may be used for AI training without taking responsibility for every downstream decision made by outside AI companies. Creators get new settings, rights holders identified through Content ID remain part of the approval process, and unauthorized scraping remains banned under existing terms.