YouTube rolls out tools to protect creators from AI voice clones, deepfakes

tech.hindustantimes.com:

YouTube plans to introduce a new feature to give content creators enhanced control over AI-generated material that mimics their voices or likenesses. The company is developing technology aimed at detecting and managing synthetic voices that replicate creators' vocal profiles. This initiative will debut with a pilot program in early 2025, according to a September 5 blog post from YouTube. The new tool, known as the synthetic singing ID, will integrate with Content ID, a system launched in 2007 that helps artists and creators manage and monetize their copyrighted content.

In addition to voice detection, YouTube is working on a second tool to help creators- including actors, musicians, and athletes- identify and handle AI-generated content that features their faces. These new tools reflect YouTube's broader efforts to address challenges associated with AI content.

Also read: Samsung tries to mock Apple's design, gets taste of its own medicine

Recently, social media platforms like YouTube have updated their policies to combat specific types of AI-generated content, including deepfakes and AI-generated voice clones. However, as these platforms advance AI-driven content creation tools, such as YouTube's Dream Screen for Shorts, they also face mounting scrutiny from intellectual property (IP) holders. These stakeholders are concerned about the use of copyrighted material to train generative AI models.

YouTube has stated that it utilises content uploaded by creators to enhance its recommendation algorithms and develop new AI features like auto dubbing. To support creators and prevent misuse of its AI tools, YouTube has incorporated safeguards designed to block prompts that infringe on its policies or cover sensitive topics.

Also read: IIT graduate joined Apple as product manager, now the key executive behind the new iPhone 16

Regarding unauthorised access to creator content by third parties, YouTube emphasised that such activities breach its Terms of Service. The platform is