YouTube will label videos with AI and combat deepfake in elections

The video platform YouTube detailed how the service will approach content made by artificial intelligence (AI). The guidelines were revealed today (06) in a post on the company's official blog.

According to the text, which was signed by CEO Neal Mohan, YouTube recognizes the “potential to stimulate creativity” of generative AI. On the other hand, the company knows its “responsibility to protect the community” and will demand identification of clips that were made or modified with digital elements.

In the coming months, YouTube will implement a tagging system for content created or edited with AI. The creators themselves will have to declare that the video is synthetic, that is, it was made with this type of digital resource.

AI videos are allowed, as long as they are flagged.AI videos are allowed, as long as they are flagged.Source: GettyImages

The brand applies to works made from scratch as well as face modifications or actions of real people — deepfakes.

In addition to the label, which may appear in duplicate in videos with more sensitive content, content that violates community guidelines will be removed. This includes, for example, the portrayal of realistic violence or misinformation.

The resources arrive in time to be implemented in at least two major electoral processes in 2024: the presidential elections in the United States, already in the primary period and scheduled for November 5th, and the municipal elections in Brazil, with the first round scheduled for November 6th. of October.

“We also know that bad faith users will inevitably try to circumvent these protection barriers and that some may succeed. We will incorporate user feedback and learning to continually improve our protections”, says the platform, already anticipating attempts to circumvent the new mechanisms.

More YouTube rules involving AI

Apart from the labels and the ban on certain content, YouTube announced that will provide communication channels for people to request the removal of content generated by AI “that simulates an identifiable person”, whether famous or not.

Additionally, record labels and copyright holders can request the removal of imitations of an artist's voice that does not have the appropriate authorization for use.

YouTube itself confirmed that it is using AI internally in some processes. In this case, the company adopted built-in functions to enhance “speed and accuracy” in content moderation systems.

Leave a Reply

Your email address will not be published. Required fields are marked *