MOSCOW, RUSSIA / RankWire.AI / – Russia’s Federation Council has endorsed a legislative bill on July 17 that establishes national guidelines for the development and deployment of large artificial intelligence foundation models. The legislation defines legal terms, grants regulatory authority, and offers support for domestically based developers meeting certain criteria. It also stipulates standards for model oversight, data storage, user notices, and AI-generated content. Having passed the State Duma on July 8, the bill now awaits presidential approval before becoming law.
According to the bill, a large foundation model is software capable of performing a wide range of intellectual tasks at a level comparable to human performance. The definition includes systems that deliver information, make decisions, or predict outcomes based on human-defined objectives. To qualify, a model must have at least 1 billion parameters. The framework further emphasizes principles related to technological sovereignty, human rights, personal choice, security, and adherence to Russian legislation.
The legislation establishes distinct categories for sovereign and national models. A sovereign model must be developed by a Russian legal entity and remain under domestic control, utilizing data centers located within Russia. Developers are required to maintain the capability to reproduce the entire development cycle, including training and original parameters. A national model follows similar ownership and data processing rules but may incorporate foreign components licensed openly.
Official recognition for domestic models
The government can implement support measures for companies involved in creating, deploying, or managing qualifying foundation models. Such measures could include access to state-held data necessary for training purposes. The law also authorizes authorities to designate specific operational environments where only sovereign or national models are permitted, especially in government information systems and other sensitive sectors. Additional regulations and presidential decrees might specify rules concerning defense, security, public order, and property security.
Websites, applications, and social media platforms are subject to a new content regulation. Platforms with more than 500,000 daily users must provide a feature allowing users to mark AI-generated audio or visual content. This rule enables users to add an informational notice to qualifying material. The format of these labels will be determined by service agreements between developers and users. While automatic labeling of all content is not mandatory, platforms must offer the technical option to do so.
Guidelines for copyright and enforcement
AI service providers are required to inform users about the ownership rights of generated outputs. They must also clarify access conditions and whether users can transfer the content. The legislation separately addresses copyrighted works used in machine learning, stating that analysis for extraction, comparison, classification, and pattern detection is lawful when developers have legally obtained access. Training on protected works is permissible if no technical restrictions were bypassed to access the material.
Most of these provisions are set to come into effect on September 1, 2026, pending presidential signature and publication. Regulations concerning domestic model designation, developer responsibilities, content marking, and intellectual property will commence on March 1, 2027. Existing systems are granted a transition period until September 1, 2032, during which their data processing and storage must remain within Russia. Until the measure is officially enacted following presidential approval and publication, it remains an approved bill, not yet a federal law.