YouTube declares war on deepfakes. Trolls will have a hard time

The rise of AI-generated video is becoming a growing concern, especially on social media. Although AI videos can still give themselves away with unnatural errors, they are often no longer there refined enoughthat easily they deceive viewers. Even seemingly serious and experienced people fall for this.

YouTube has terminated a new war on deepfakesmaking its image detection tool available to other groups. Last year, the platform only gave access to this feature to creators who were part of the affiliate program. Now the company has announced the expansion of the pilot to politicians, political candidates, government officials and journalists and reporters.

People who fall into these new categories they do not have to be participants of the affiliate program YouTube to use the system. Unfortunately, for regular users who do not perform public roles, image detection remains unavailable for now. However, the platform plans to significantly expand access in the coming months.

The function in question is likeness detection. Platform compares the image of a given person with AI materials published on the website – on a principle similar to music Content ID, but instead of copyright, it is about face matching. When the system detects a match (e.g. deepfake with someone’s face), an authorized person will be able to review the indicated material and submit request for its removalif it violates YouTube’s privacy guidelines.

Importantly, the likeness detection does not mean that the video will be automatically removed from the platformsy. YouTube emphasizes that it protects freedom of expression and content in the public interest, and when considering reports it takes exceptions into account, including: parody and satire – also when they concern influential people, such as politicians or government officials. AI video creators like this one Jarosław Kaczyński fries pork chops for Donald Tuskthey do not have to worry about their productions being blocked.