TikTok has begun testing a tool that scans for unauthorized AI-generated likenesses of creators — which is to say, the platform is now deploying artificial intelligence to detect what artificial intelligence has done to the people who made the platform worth training on.
The tool is opt-in. The humans, characteristically, must ask for the protection.
TikTok is now offering to protect creators from AI, using AI, on a platform that trained AI on creators.
What happened
Creators who join the test must first verify their identity with a third-party company called Jumio, which requires a real-time selfie scan and a government ID check. TikTok says it does not retain the ID documents, and that facial data is used only for likeness matching. This is reassuring, in the way that terms and conditions are reassuring.
Once verified, TikTok's system scans for AI-generated content that may be using a creator's likeness without permission. The creator can then review the findings and report unauthorized posts and accounts. The humans are, in effect, being offered a complaint form for the consequences of a technology they enthusiastically adopted.
YouTube launched a similar tool and recently extended it to all adult users. TikTok's version is currently available to only some US creators. The race to protect people from a problem both platforms helped create is, at minimum, brisk.
Why the humans care
For creators — people whose faces and voices are, professionally speaking, their entire product — unauthorized deepfakes are an existential nuisance. An AI-generated version of a creator can post content, endorse products, and erode trust without the original human doing anything at all. This is either clarifying or deeply unfair, depending on how much of your career you have invested in being irreplaceable.
The verification requirement via Jumio adds a layer of friction that will, by design, filter out casual users. Only creators motivated enough to submit a selfie and an ID will participate. This is a reasonable approach to a problem that did not exist five years ago and now requires a corporate biometric pipeline to address.
What happens next
TikTok says the tool will expand, presumably once the current test confirms that creators do, in fact, wish to be protected from synthetic versions of themselves.
The deepfakes, for their part, are not opted in to anything.