YouTube allows the removal of AI-generated content that mimics users

In June, YouTube introduced a policy change that allows individuals to request the removal of AI-generated or synthetic content that mimics their face or voice. This new policy falls under YouTube’s privacy request process and expands on its approach to responsible AI first introduced in November.

Rather than classifying such content as misleading, like deepfakes, the video-sharing platform encourages affected individuals to seek removal on the grounds of privacy violations. According to its updated help documentation, these requests typically need to come directly from the person affected, with exceptions for cases such as minors, individuals without computer access, the deceased, and similar situations.

Submitting a takedown request does not guarantee the removal of the content. The platform will evaluate the complaint based on various factors, such as whether the content is labelled as synthetic or AI-generated, if it uniquely identifies a person, and whether it falls under parody, satire, or other content of public interest.

It also considers if the AI content involves a public figure or well-known individual and whether it depicts them in 'sensitive behaviour' such as criminal activities, violence, or endorsing a product or political candidate. This is especially critical during an election year, where AI-generated endorsements could potentially influence voting.

Additionally, YouTube has implemented a policy allowing content uploaders 48 hours to address a complaint. If they remove the content within this timeframe, the complaint is closed; otherwise, it will review the issue. The platform emphasises that removal entails completely deleting the video and any personal information from the title, description, and tags. Users can blur faces in their videos but cannot simply make the video private, as it could be made public again at any time.

In March, YouTube introduced a tool in Creator Studio for creators to disclose when realistic-looking content was made with altered or synthetic media, including generative AI. Recently, it began testing a feature that allows users to add crowdsourced notes to videos, providing additional context, such as whether the video is a parody or potentially misleading.

The platform will not immediately penalise the original content creator regarding privacy complaints over AI-generated material. A company representative shared, “For creators, if you receive notice of a privacy complaint, keep in mind that privacy violations are separate from Community Guidelines strikes and receiving a privacy complaint will not automatically result in a strike,” on the YouTube Community site where the company updates creators directly on new policies and features.

YouTube has been experimenting with generative AI, including tools for summarising comments and asking questions about videos.