From March 1, 2027, large platforms with a Russian audience of more than 500 thousand people per day will be obliged to provide users with the opportunity to mark content created with the help of AI. This refers to audio and video. 

Social networks will not be required to detect neurocontent on their own — the mechanism is designed for users. Authors will be able to add tags themselves. For example, a checkbox, entering data into a separate field when uploading, or marking AI in the publication interface. From a technical point of view, implementing such a tool is no more difficult than the usual age rating.

However, there is also a more protected level — the international C2PA standard. It attaches cryptographically verified information to an image, audio, or video about how and when the file was created and what changes were made to it.

At the same time, no technology provides an absolute guarantee, experts believe. According to them, metadata can be erased when copying or the file can be reassembled, and an unscrupulous distributor of a deepfake is unlikely to voluntarily put a tag.

The first Russian AI law

This refers to Federal Law No. 243-FZ "On the Development, Implementation, and Application of Large Foundational Artificial Intelligence Models in Russia".

The document for the first time in the Russian Federation creates a general legal framework for large foundational AI models with at least 1 billion parameters. It introduces the concepts of "sovereign" and "national" models, provides for state support measures for developers, and establishes rules for the use of intellectual property results in model training.

Read more about the law in the material "Pervyy Tekhnicheskiy".