Cut.ProCut.Pro

TikTok launches a public AI detector and a tool against deepfakes

TikTok opened a page where anyone can check whether a photo or video came out of its AI tools, and it took face detection for deepfakes worldwide. Here is what that changes for people who post AI-made video and for anyone using another person's face or voice.

TikTok launches a public AI detector and a tool against deepfakes

TikTok launches a public AI detector and a tool against deepfakes

On Friday (Nov. 20), TikTok announced two tools that affect anyone who posts AI-made video: a public checker that tells you whether a photo or video came out of TikTok's own AI tools, and the global expansion of Likeness Detection, which alerts a creator when their face shows up in a deepfake. Hiding where a generated video came from just got harder, and using someone else's face just got riskier.

The information comes from Mobile Time, a Brazilian outlet, and from AFP, republished by Free Malaysia Today.

What AI Detection is

It is a page in TikTok's Transparency Center where anyone, even without an account, uploads a photo or video and finds out whether it was created or edited with the platform's AI tools. According to Mobile Time, the check reads two things:

  • Content credentials (C2PA): a record written into the file saying how it was made. It is an open standard that other platforms read too.
  • Invisible watermark, which Mobile Time calls VeraDura: a signal embedded in the image that viewers can't see.

The limit is the scope. The tool recognizes what came out of TikTok's own AI tools, not any AI video on the internet. A video generated somewhere else, without credentials, won't flag anything there.

The platform says it has labeled more than 5 billion videos as AI, again according to Mobile Time, and the hub it opened in July to teach people how to spot generated content passed 4 million visits.

What Likeness Detection is

It is an opt-in tool for eligible creators and public figures, including government accounts, politicians and parties. The person signs up, confirms their identity with a partner company and starts receiving alerts when their face appears in AI-generated content without permission. Then they decide whether to report it.

TikTok had been testing this since July with only a small group of creators in the United States. At the time, spokesperson Zachary Kizer told The Verge that verification uses Jumio, with a live selfie and an ID document, that TikTok does not keep the document, and that face data is used only to find matches, according to Digital Trends.

Now the expansion is global and gradual. The stated reason is elections: according to AFP, TikTok ties the tool to the 2027 European elections, such as those in Italy, France and Greece, and Francesca Scapolo, who leads government relations for European elections, said at an event in London that protecting election integrity is one of the company's most critical responsibilities. YouTube already has a similar tool that alerts creators when they appear in AI videos that imitate them.

Does it apply in Brazil?

The launch is global, according to Mobile Time, so the public checker should open in the Brazilian market too. Likeness Detection is a different story: it rolls out gradually, only for people TikTok considers eligible, and none of the sources say what the criteria are or when a Brazilian creator gets the invitation.

What changes for people who post AI video

The labeling rule stays the same, but it got easier to check who follows it.

TikTok's guidelines already require a label on content created or heavily edited with AI that shows realistic people or scenes, and they say an unlabeled video "may be removed, restricted or labeled" by the team. Since 2024, TikTok has also labeled automatically anything that arrives from other platforms with C2PA credentials.

You generate video with TikTok's tools and repost it elsewhere. The file carries the mark. Anyone, including the brand that hired you or a suspicious follower, can now check it in seconds.

You make a persona or avatar with your own face. Nothing changes except the label, which you should already be turning on. If you are still building yours, we cover the options in AI persona: how to create one that looks like you.

You use another person's face. Politician parodies, a streamer saying things they never said, a celebrity endorsing a product. That is the format the tool targets. If the person is signed up, they get your finished video ready to report. We already saw how this ends on YouTube with the Draco Malfoy AI channels. I wouldn't build a channel on it right now.

It is the same move as Instagram, which this week launched Content Protection and will test detection of AI video that reuses another creator's audio.

What about parental controls

The package also brought a revamped Family Pairing, with a browser version. The dashboard shows parents screen time, follower changes and blocks, a sample of who the teen follows, and the public posts from the past week with the views on each. A new alert tells parents when the teen searches self-harm terms several times. If you have a teen audience or team, this adds to the AI age check that Instagram rolled out worldwide.

What to do now

  1. Turn on the AI label for every realistic generated video. If you publish to TikTok through Cut.Pro, the network form has a "Generated by AI" switch: turn it on when the video shows a realistic person or scene made with AI.
  2. Test your own file in the checker when it opens in Brazil, so you know what the platform sees before someone checks for you.
  3. Take third-party faces and voices out of your script, unless you have written permission.
  4. If you are the one being copied, watch for the Likeness Detection invitation and for Instagram's Content Protection. If you are imitated often, you finally get a way to find the deepfake without depending on a follower sending you a screenshot.
Share

Keep reading

More insights and tutorials to help you grow as a content creator.