⚖️ RAO May Be Entrusted with Managing Opt-Outs for AI Training on Copyrighted Content

Russia is preparing a new mechanism to protect intellectual property from unauthorized use in training artificial intelligence models.

Under amendments to the law “On Supporting AI Development in the Russian Federation,” drafted by a working group under the Presidential Council for Codification, authors may be required to publish notices barring neural networks from processing their content on the platform of a collective management organization (CMO).

This new approach pivots away from an earlier concept that relied on embedding metadata directly into the creative works or author websites. Media industry insiders expect the Russian Authors’ Society (RAO) to be granted sole CMO status. This would enable copyright holders to claim lump-sum compensation from AI developers for prior training conducted on their materials. The Ministry of Digital Development and the office of Deputy Prime Minister Dmitry Grigorenko confirmed that regulatory refinement remains ongoing in consultation with industry stakeholders.

Key points of the initiative and industry reactions:

  • Registry Centralization. Replacing file-by-file tagging with a unified database simplifies administrative tracking for rights holders. However, delegating these duties to a non-governmental entity like RAO has raised concerns over the risk of creating a commercial monopoly.
  • Music Industry Position. The National Federation of the Music Industry (NFMI) views a unified registry as technically sound, but insists it should be managed by a government agency rather than a commercial CMO to ensure regulatory transparency.
  • IT Sector Criticism. The Big Data Association—which includes Yandex, Sber, and Avito—calls an independent centralized registry technically infeasible. Cross-referencing billions of dynamic objects against an external database prior to each training cycle would severely bottleneck the development of domestic AI models.

Source: Kommersant