As AI-generated content proliferates, numerous companies and platforms are grappling with how to manage it. The labeling of such media to make it obvious that an AI model created a picture, video, or audio clip is a major worry for many. To do this, YouTube has published new guidelines for tagging videos created using AI.
In the interest of transparency, the site will now demand that anyone posting a realistic-looking video that “is made with altered or synthetic media, including generative AI” provide a disclaimer. Anything that a viewer could “easily mistake” for an actual person, event, or location is considered realistic content according to YouTube.
A creator must add a label if they substitute someone else’s face for another person’s or use a synthetic rendition of a real person’s voice to narrate a video. If they adjust the video of an actual incident or location (e.g., by altering an existing cityscape or making it appear as though a genuine structure is on fire), they will also have to provide the disclosure.
If a video creator hasn’t applied one of these labels, YouTube says it may do so “especially if the altered or synthetic content has the potential to confuse or mislead people.” The team points out that although it wants to allow creators some time to adjust to the new guidelines, those who consistently break the rules by not including a label when they should probably face consequences from YouTube.
In the upcoming weeks, these labels will begin to show up on YouTube, first on the mobile app and later on desktop and TV screens. The expanded description, which states that the video contains “altered or synthetic content” and that “sound or visuals were significantly edited or digitally generated,” is where they will mostly be found.
But YouTube will prominently label more delicate subjects (including news, politics, money, and health) by placing the label right on the video player.
If creators simply employed generative AI to assist with things like writing scripts, coming up with ideas for movies, or automatically creating subtitles, they wouldn’t need to include the label. For “clearly unrealistic content” or if alterations have little impact, labels won’t be required. Creators will not be required to utilize the altered content label when making color adjustments or applying special effects like adding background blur on their own. Applying beauty, lighting, or other filters won’t help either.
Furthermore, YouTube reports that it is still in the process of redesigning the takedown request procedure for the modified or synthetic video that has the voice or face of a genuine, distinguishable person. It intends to release additional information about that revised process shortly.
