A deepfake can spread across social networks faster than a person or company can prove that everything in the video, photo, or audio is falsified. Therefore, the Main Radio Frequency Center of Roskomnadzor proposed that platforms themselves identify suspicious materials and restrict their distribution even before verifying the information with those mentioned in them. The initiative was announced at a forum dedicated to trusted artificial intelligence technologies.
Roskomnadzor believes that clear rapid response mechanisms are needed to regulate deepfakes. The idea is that social networks could temporarily stop the spread of suspicious content until its authenticity is confirmed.
Among additional measures, the labeling of generated materials and the creation of a database of technological threats are also being discussed. At the same time, market participants note that social networks do not yet have a reliable tool that would accurately identify all AI content.
According to industry representatives, modern generative models quickly bypass detectors. In addition, erroneous blockages can lead to conflicts with users and brands.
Experts believe that a hybrid scheme may be the most realistic option. First, automatic filters will weed out suspicious materials, and then the most controversial cases will have to be checked manually.
According to experts, it will be difficult to fully automate such a system. But it is possible to partially launch it based on existing solutions.