Russian legislation in the field of artificial intelligence is preparing for a serious transformation. According to a draft of amendments to the recently adopted law on AI and the Civil Code, labeling of images, audio, and video created using large language models could become mandatory. This fundamentally changes the approach laid out in the current version, where such measures were purely voluntary.
The initiative is in the active discussion stage. The document has been prepared by a working group of the Presidential Council of the Russian Federation on the codification of civil legislation, led by the chairman of the State Duma committee on state construction, Pavel Krasheninnikov. It is planned that the key amendments will be reviewed as early as August 19 at a platform overseen by State Secretary and Deputy Minister of Digital Development Ivan Lebedev.
What exactly is changing?
The key novelty is the establishment of an obligation to label content created "using" neural networks. This wording is extremely broad and could even cover materials that have merely been edited with the help of AI. At the same time, no threshold for the minimum share of AI generation has been set in the project, which creates grounds for legal uncertainty.
The question of responsibility also remains open: who should place the label if content is created by one person and distributed by another? The project does not introduce special liability for the absence of labeling and does not define the body that will monitor compliance with the norm. It also does not specify whether the label must "survive" the transfer of a file between platforms. Large websites and social networks with an audience of more than 500,000 Russian users per day, according to the current version, are only required to ensure the technical capability for applying such a mark.
The timeline for the framework provisions to take effect is set for March 1, 2027, but the exact dates for submitting the amendments to the State Duma have not yet been determined.
Arguments against and the international context
In the expert community, the initiative raises serious objections. The main thesis of critics: mandatory labeling will not become an effective tool in the fight against deepfakes. Malicious actors will continue to use foreign services and distribute materials bypassing any requirements. Moreover, the norm will create additional barriers for legitimate Russian AI services without solving the root problem.
Technical unpreparedness also draws attention. As experts rightly note, the world has still not resolved the question of how to reliably preserve a machine-readable label under any manipulation of content. Introducing mandatory requirements before the technology is refined looks premature.
Foreign experience is also noteworthy. The European AI Act requires providers to ensure machine-readable labeling and distributors of deepfakes to disclose their origin. China has gone even further by implementing a hybrid model: services place both visible and hidden labels, platforms verify them, and users can report the synthetic origin of content. Russian regulation, it seems, may move toward exactly such a mixed scheme.
My analysis: For now, the project remains raw. Shifting all significant decisions to by-laws and the absence of clear technical standards is a path to chaos in the market. Instead of rushing with mandatory norms, the regulator should first launch pilot projects and refine the mechanism on a voluntary basis. Otherwise, we risk getting a law that cannot protect users but will create excessive administrative burden on business.