YouTube quietly updated its content removal policy to address AI-generated content. Users can now request takedowns of synthetic content that imitates their face or voice through the platform’s privacy request process. This policy expansion builds upon YouTube’s responsible AI initiative (launched in November), reflecting the growing concerns surrounding artificial intelligence and content authenticity.
I cut and pasted the relevant portion of YouTube’s Privacy Complaint Process below. Pay careful attention to the wording. As you read it, imagine that you are the adjudicator of the complaint. Also, try to imagine if you were the complainant. Then, try to imagine yourself as the creator who posted the offending content. Did you actually mean to violate this policy? As a creator, are the guidelines clear?
Reporting a privacy violation
If someone has used your personal information without your permission, here are ways to report it:
Report AI-generated or other synthetic content that looks or sounds like you:
If someone has used AI to alter or create synthetic content that looks or sounds like you, you can ask for it to be removed. In order to qualify for removal, the content should depict a realistic altered or synthetic version of your likeness. We will consider a variety of factors when evaluating the complaint, such as:
- Whether the content is altered or synthetic
- Whether the content is disclosed to viewers as altered or synthetic
- Whether the person can be uniquely identified
- Whether the content is realistic
- Whether the content contains parody, satire or other public interest value
- Whether the content features a public figure or well-known individual engaging in a sensitive behavior such as criminal activity, violence, or endorsing a product or political candidate
To start a report, follow the Privacy Complaint Process for altered or synthetic content.
I appreciate where Google is trying to go with this. It makes sense to start the process with aggrieved individuals. I know you should never say never, but there will never be a way to automate this process, so – in practice, at YouTube scale – this is an intractable problem. I could be wrong about this… but I’m not.