YouTube has announced that authors can independently label videos that contain artificial or artificially generated content.
The corresponding checkbox appears when uploading and publishing a video. Authors should indicate «altered or synthetic» content that looks realistic. This covers things like making a real person say or do something they didn’t say or do; changing footage of real events and places; showing a «realistic scene» that didn’t actually happen. Some examples that YouTube offers include showing a fake tornado moving toward a real city or using dipshots to make a real person narrate a video.
However, you don’t need to check the boxes for things like beauty filters, special effects like background blur, and «obviously unrealistic content» like animation.
Like other platforms that have implemented AI content labels, YouTube’s feature relies on an honor system — creators must be honest about what appears in their videos. YouTube spokesperson Jack Malone has previously said that the company «invests in tools» to detect AI-generated content, although AI detection software has historically been very inaccurate.
YouTube also states that it can add artificial intelligence disclosures to a video even if the user who uploaded it did not do so. This is especially true if the altered or synthetic content is likely to confuse or mislead people «. More prominent marks will also appear on the video itself for sensitive topics such as health, elections, and finance.
Source: The Verge