YouTube Now Requires Creators to Label Videos Created Using Generative AI
Is anyone else worried about the future of the Internet? I know it does, especially now that AI-generated realistic videos are becoming increasingly popular . I imagine it won’t be long before even the smartest among us are baffled by an image that doesn’t actually exist. Any steps we can take now, before it’s too late, to make it easier to identify what content is AI-generated will be steps in the right direction.
Today, YouTube is taking one of those steps: On Tuesday, the company announced that it will be rolling out a new tool in Creator Studio that will require creators to flag whenever realistic content is created using “altered or synthetic media.” Of course, this includes media created using generative artificial intelligence. You’ll find this option in the new Modified Content section, where you can select “Yes ” or “No” depending on whether your video meets the requirements.
YouTube will not require creators to flag videos that use obviously altered or synthetic materials, including animation or videos with special effects. However, the following categories will require authors to use the new label:
-
If you use someone’s image in your video, say realistically replace one person’s face with another (deepfake).
-
If you create a voice based on the voice of a real person.
-
If you change a real place in a realistic way, such as setting a real building on fire or editing a real skyline.
-
If you create a realistic scene that the audience is tricked into believing is real, such as “a tornado heading towards a real city.”
To be clear, this label is only required for the visual and audio elements of this download. YouTube will not require creators to flag when they have used generative AI for other areas of video creation, such as scriptwriting, nor will there be a need for a label if the video is not realistic. This may include animation, lighting and color effects, certain special effects that are not realistic or substantial, or certain facial filters.
Unfortunately, this will likely not be the transformative change that many of us concerned about AI manipulation will be expecting. Although YouTube “requires” creators to add this label whenever their content meets these requirements, creators can still upload videos without applying the label. YouTube says it will “consider enforcement action” against creators who don’t properly label their videos, and perhaps even apply the labels themselves, but there is no clear policy in place yet. At this point, this labeling system is largely based on the honor system.
However, we hope creators are playing the game here and labeling their videos accordingly. YouTube says the tag will appear in the rich description of most videos, but for those that are particularly sensitive (YouTube covers health, news, elections or finance), the site may apply a more obvious tag to the video itself.