The short video platform TikTok has launched beta testing of a new tool, Likeness Detection, designed to protect copyrighted content from misuse by neural networks. The feature, currently available to a limited group of American creators, automatically detects AI-generated video material that uses a specific blogger's face.
The tool's mechanism is based on biometric analysis. To activate it, users must verify their identity through a third-party service, Jumio, by providing a real-time selfie and a scan of an identification document. After confirmation, the system begins scanning the platform for deepfakes and fake accounts that mimic the author's appearance.
Protection Against AI Copying: How It Works
Upon detecting suspicious content, the blogger can promptly file a complaint for violation. TikTok emphasizes that it does not store copies of uploaded documents, and biometric data is used solely for facial matching and does not leave the secure environment. This is a crucial step in combating the growing threat of digital fraud, where attackers create realistic AI copies of celebrities to spread fakes or scams.
From my perspective, the implementation of Likeness Detection is a timely and technically mature response to the challenges of the generative AI era. However, the solution's effectiveness directly depends on the scale of the verified user base. Currently, the tool covers only a small portion of the audience, leaving loopholes for attackers. In the long term, such mechanisms will become a mandatory standard for all major platforms, as the cost of reputational damage from uncontrolled deepfakes already exceeds the expense of implementing biometric protection.