YouTube plans to ask creators to flag realistic videos and audio that have been generated or altered with AI. The platform says viewers will see labels explaining when content is synthetic, while creators who do not disclose qualifying material could face penalties.
Creators will be asked to flag realistic AI content
The disclosure changes are expected to roll out over the coming months and into next year. During upload, creators will have new options to indicate whether their video includes realistic AI-generated or AI-altered material.
The rule targets content that could be mistaken for a real event or a real person’s actions. YouTube gave the examples of a video that depicts an event that never happened and a video showing someone saying or doing something they did not actually do.
The company says the policy is meant to address risks from generative AI, including deepfakes, voice cloning and disinformation. YouTube already has policies against technically manipulated content that misleads viewers, but it says newer AI tools can make altered or synthetic content harder to recognize.
Labels will give viewers more context
YouTube plans a labeling system to tell viewers when content has been altered or is synthetic. Labels are expected to appear in the description panel and video player, with particular attention to videos about sensitive subjects such as elections, ongoing conflicts, public health crises and public officials.
Videos made with YouTube’s own generative AI products, including its AI-powered video creator Dream Screen, will be labeled automatically. The company shared mock-ups of possible labels and said their appearance could change over time.
YouTube also plans to use AI-powered moderation tools to help identify and handle content that breaks the new rules. Creators who do not disclose AI use may face consequences that include removal of their content or suspension from the YouTube Partner Program.
People and artists may request removals
The platform says individuals will be able to ask for AI-generated content that simulates their identifiable face or voice to be removed through a privacy request process. That could include deepfakes, but a request will not guarantee removal.
YouTube says it will consider the circumstances of each request. Factors may include whether the content is parody or satire, whether the person can be uniquely identified, and whether the subject is a public official or well-known individual. The company says those cases may face a higher bar for removal.
Artists and music publishers will also be able to request removal of AI-generated music that imitates an artist’s distinctive singing or rapping voice. YouTube says it will consider whether the material is used for news reporting, analysis or critique of the synthetic vocals.
Rules will have to account for creative uses
YouTube describes the changes as an effort to protect the information environment while making room for new creative uses of generative AI. Its stated approach recognizes that synthetic content may be misleading in one context but used for parody, fair use or political commentary in another.
The disclosure and labeling system gives viewers a way to understand how a video was made, while removal requests offer a route for people concerned about likeness or voice imitation. How the policies work in practice will depend on how YouTube applies those factors as the measures roll out.