The system recognizes content created by modern AI models, such as Veo 3 (Google) and Sora 2 (OpenAI), with an accuracy of over 98%.
Currently, MTS Web Services is testing a new deepfake detector with one of the state services in Russia, three banks in Russia and the CIS, the MTS Link video conferencing and online learning platform, and the MTS Defender service, which warns users about conversations with potential scammers.
Deepfakes are recognized by analyzing the audio track using a detector from MWS AI and analyzing images and videos using a detector from VisionLabs. In the audio part, a specialized model is used: first, it is trained on raw recordings of human speech to "understand" natural sound patterns, and then it is fine-tuned on synthetic recordings, identifying characteristic features of the generated voice. MTS explained:
The high accuracy of recognition of videos created by modern AI models allows you to automatically identify fake videos before publication, filter AI content in media, social networks and messengers, prevent the spread of videos with fake images of politicians and public figures, and counteract attempts at mass manipulation, for example, when creating fake videos of disasters and emergencies that can cause panic.
