Authors will be able to prohibit the use of their works for training neural networks through a unified system. According to "Kommersant", the copyright holder will only need to place such a prohibition on their website or in a special database of a rights management organization. Currently, it is proposed to place the restriction mark directly on the work itself. Amendments to the law on artificial intelligence are being developed by a working group under the Council for Codification and Improvement of Civil Legislation under the President of the Russian Federation.

The current version of the law provides for placing a prohibition mark directly on the work. The new version of the amendments proposes to change this approach. The author will be able to place the prohibition on their website or on the official resource of a rights management organization.

Two "Kommersant" sources in major media holdings believe that the Russian Authors' Society (RAO) could become such an organization. They do not rule out that in the future, RAO could act as an intermediary in collecting payments from AI model developers.

Developers, meanwhile, are expected to be required to pay compensation to copyright holders once for AI training that has already been conducted on their content. "Vedomosti" reported this on August 17, citing a document. RAO did not respond to "Kommersant"'s request.

The Ministry of Digital Development stated that the initiative is being worked out together with industry participants. The department considers it important to maintain the approaches of already adopted legislation, including the ability of copyright holders to protect their rights using technical means. The office of Deputy Prime Minister Dmitry Grigorenko reported that various issues of AI use in relation to copyright objects continue to be discussed with businesses.

The idea of a unified database itself is evaluated differently. Nikita Danilov, CEO of the National Federation of Music Industry, considers the rejection of marks on each work understandable from a technical point of view. Marking each individual file is practically impossible, and it is difficult for an AI developer to track such marks on thousands of objects. Therefore, a centralized registry is easier to administer.

The Association of Big Data, which includes Avito, Sber, Yandex, and other companies, holds a different position. They believe that a separate centralized database will only complicate the system and will not be technically able to work. Modern AI models use huge arrays of constantly changing data, so a developer cannot compare billions of objects with a separate registry of prohibitions before each training.

A media market representative also noted that it is easier to work with a clear machine-readable prohibition directly at the source. Otherwise, AI training could turn into a constant bureaucratic check.