TikTok will flag users who upload artificial intelligence-generated content (AIGC) to the video-sharing site from other platforms, the company says, becoming the first major video site to automatically tag such content for users to see.
Content created with TikTok’s own AI tools is already automatically flagged as such for viewers, and the company has required creators to manually add the same tags to their own content, but they have so far been able to evade the rules and pass off the generated material as authentic by uploading it from other platforms.
Now, the company will begin using digital watermarks created by the cross-industry group Coalition for Content Provenance and Authenticity (C2PA) to identify and label as many AIGCs as it can.
“AI offers incredible creative opportunities, but it can confuse or mislead viewers if they don’t know the content was generated by AI,” said Adam Presser, head of operations and trust and safety at TikTok. “Labeling helps clarify that context, which is why we label AIGC made with AI effects from TikTokand have required creators to label AIGC realistic for over a year.”
Labeling goes both ways: TikTok will also begin applying the same digital watermarking technology, called Content Credentials, to content downloaded from its own platform, allowing other platforms to identify “when, where and how it was created or edited the content.” Presser said.
But the ability to tag generated content is limited to that created by other platforms that are also members of C2PA. That includes most of the major AI players, such as Microsoft, Google, and Adobe.
Until this week, it did not count OpenAI among its members, but the research lab joined the steering committee on Tuesday. It had already started using Content Credentials technology earlier this year and plans to include it in its video creation AI, Sora, once it is released to the public.
But smaller, less scrupulous and less commercially focused AI groups will continue to produce unlabeled content for some time. Open source tools like Stable Diffusion, whose underlying code can be downloaded for free, can always be modified to remove any attempts to tag images (although Stability.AI, one of the creators of Stable Diffusion and current developer of the tool, is a member of the group). AI startup Midjourney, one of the most popular imaging tools, is completely absent from the membership list.
TikTok’s tagging plans follow Meta, which made a similar announcement in February. But attempts at labeling have done little to curb the huge proliferation of AI-generated images on Facebook in particular. leading some to call the site the “Zombie Internet.”.
Other social media apps have been slower to get it right: Snapchat labels AI-generated content created with its own tools, but warns users that “images created with non-Snap products cannot be labeled as AI-generated,” while Elon Musk’s