TikTok announced Thursday that it will begin using technology generated by artificial intelligence and aimed at helping label images and videos uploaded to the video-sharing service.
TikTok has announced that it will adopt “Content Credentials,” digital watermarks that show how images were created and edited. Although content credential technology was pioneered by Adobe ADBE.O, it is available to other companies and has already been adopted by companies such as OpenAI, which created ChatGPT.
Researchers have expressed concern that AI-generated content could be used as misinformation to interfere with this fall's U.S. elections. TikTok was already part of a group of 20 tech companies that signed an agreement pledging to fight this earlier this year.
Alphabet GOOGL.O Google's YouTube, as well as Metaplatforms META.O, which owns Instagram and Facebook, have also announced plans to use content authentication information.
For the system to work, both the makers of the generative AI tools used to create the content and the platforms used to distribute the content must agree to the use of industry standards.
For example, when a user generates an image using OpenAI's Dall-E tool, OpenAI attaches a watermark to the generated image. When a marked image is uploaded to her TikTok, it will automatically be labeled as her AI-generated.
TikTok, owned by China's ByteDance, has 170 million users in the United States, and recently passed a law requiring it to sell TikTok to ByteDance or face a ban. . TikTok and ByteDance have filed suit to block the law, claiming it violates the First Amendment.
TikTok already labels AI-generated content created using tools within the app, but the latest move means the labels will also apply to content generated outside of the service.
“We also have a policy that prohibits unlabeled Realistic AI, so if Realistic AI (generated content) appears on our platform, we will remove it as a violation of our Community Guidelines,” the company said. said Adam Presser, Director of Safety. TikTok said in an interview.
(Reuters)
