YouTube on Thursday announced a new set of AI detection tools to help prevent artists, actors, musicians, athletes and other creators from having their likenesses, including their faces and voices, copied and used in other videos. One key component of the new detection technology is an expansion of YouTube's existing Content ID system, which currently identifies copyrighted material. That system will be expanded to include new synthetic singing voice identification technology to identify AI content that imitates someone's singing voice. The company said other detection technologies will also be developed to identify when someone's face has been imitated with AI.
It's also worth noting that YouTube is only just beginning to come up with a solution to address the issue of its content being used to train AI models, which has been an issue for some time, with creators complaining that companies like Apple, Nvidia, Anthropic, OpenAI, and Google are training their content without their consent or compensation.
YouTube hasn't yet revealed how it plans to help protect creators (or generate its own additional revenue from AI training), saying only that something is in the works.
“…we're developing new ways for YouTube creators to choose how third parties use their content on the YouTube platform, and we'll have more details to share later this year,” the announcement states succinctly.
Meanwhile, the company appears to be moving forward with a promise it made last year to come up with a way to compensate artists whose work is used in the creation of AI music. At the time, YouTube began working with Universal Music Group (UMG) and its talent on a solution. It also said it would work to expand its Content ID system to identify which rightsholders should be compensated when their works are used in AI music. YouTube notes that the Content ID system now processes billions of claims a year, generating billions of dollars in revenue for creators and artists.
YouTube didn't mention the AI music compensation element in today's announcement, but said it is close to piloting an expansion of its Content ID system focused on this area: YouTube said it will begin testing synthetic singing identification technology with partners starting early next year.
Another solution in early stages of development will allow public figures, including actors, musicians, creators and athletes, to detect and moderate AI-generated content that features their face on YouTube. This would go a long way in preventing their likeness from being used to mislead YouTube viewers, for example to promote products or services they don't agree to endorse or to spread misinformation. YouTube didn't say when the system would be ready for testing, saying it's currently in development.
“As AI evolves, we believe it should enhance human creativity, not replace it. We are committed to working with our partners to ensure future advancements amplify their voices, address their concerns, and continue to develop guardrails to achieve our shared goals,” YouTube's announcement said.