MOSCOW, RUSSIA / RankWire.AI / – The Federation Council sanctioned a legislative framework targeting artificial intelligence on July 17, establishing guidelines for large foundational models in Russia. The legislation specifies the technologies it covers and grants authority to government agencies. It also sets standards related to model ownership, local data storage, user transparency, and AI-generated content. Having cleared the State Duma on July 8, the bill awaits presidential approval and formal publication before becoming law at the federal level.
The proposed law characterizes a large foundation model as software capable of executing numerous intellectual tasks at a level comparable to humans. To qualify, a system must contain at least 1 billion parameters. Such models can provide information, aid decision-making, or predict outcomes based on human-defined objectives. The framework emphasizes principles including technological sovereignty, human rights, personal choice, security, and legal compliance in Russia. These principles govern the development, deployment, and application of qualifying AI systems.
The legislation introduces classifications for sovereign and national models, linked to Russian jurisdiction. A sovereign model must originate from a Russian legal entity and operate on data centers within the country. Its creators must be able to replicate the entire development process, including training and original parameters. A national model adheres to similar ownership and localization rules but may incorporate foreign open-license software components, provided Russian entities maintain necessary control and operational capacity.
Legal Designations for Domestic AI Systems
The government might provide assistance to developers involved in creating, deploying, or managing qualifying foundation models. This aid could include access to state-controlled datasets for training purposes. Authorities may also mandate exclusive use of sovereign or national models within government information systems and other sensitive environments. Additional rules concerning defense, security, public order, and property protection could be established through separate legislation or presidential decrees. The framework assigns responsibility to state agencies for enforcing these requirements within their legal authority.
Large digital platforms face an additional mandate regarding AI-generated audiovisual content. Platforms with over 500,000 daily users must implement tools allowing users to label such material. This requirement applies to websites, apps, and social media services. It does not necessitate automatic labeling of every item; instead, developers and users can agree on the disclosure format via service agreements. The core goal is to provide an option for creators or distributors of qualifying content to disclose its AI origin.
Standards for Copyright and Content Transparency
AI providers are required to inform users about rights ownership of generated materials. They must clarify access terms and whether content can be downloaded or transferred. The legislation also addresses the use of copyrighted works in training data. It allows analysis for extraction, comparison, classification, and pattern recognition when lawful access is obtained. Training on protected works is permitted when no technical access restrictions are bypassed. These rules align model training practices with existing copyright and access regulations.
Most provisions are scheduled to take effect on September 1, 2026, subject to presidential approval and official publication. Regulations concerning domestic model classification, developer obligations, content marking, and intellectual property will be implemented starting March 1, 2027. Existing systems may operate until September 1, 2032, as long as they process and store data within Russia. Until the bill is formally signed and published, it remains an approved legislative proposal rather than an enacted law under Russia’s legislative process.