JAKARTA - YouTube is expanding its similarity detection feature on its platform to more workers in the entertainment industry. This Artificial Intelligence (AI) feature is designed to prevent digital identity abuse.
This detection capability is designed to identify deepfake content or celebrity faces uploaded without permission. The way this technology works is similar to the Content ID system used to detect copyrighted material.
The difference is that this similarity detection feature focuses on the visual match of the faces of public figures listed in the YouTube database. This step is taken to reduce the number of ad fraud that often includes the faces of celebrities or artists illegally.
YouTube wants to ensure that the identities of creators and public figures are not used for harmful purposes. Previously, this technology had been tested in a pilot program for a small number of creators last year.
However, starting this spring, the scope of access has been expanded to more parties, including journalists, politicians, and government officials. Now, talent agencies such as CAA, UTA, and WME are also joining in developing this protective tool.
Even though they don't have a personal YouTube channel, celebrities are still protected with this feature. The system will scan the content generated by AI and provide notifications if a visual match is found.
After being detected, the party who is harmed can choose to request the deletion of the video or file a privacy claim. However, YouTube still provides exceptions for content that is parody or satire according to community rules.
This is done to maintain a balance between protecting individual rights and freedom of expression. In the future, YouTube plans to expand the capabilities of this system so that it can detect audio similarities of celebrities and famous figures.
The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)
