Russia adopted its first AI law

On 26 July 2026, Russia adopted Federal Law No. 243-FZ* (the “Law”) to support the development of artificial intelligence (the “AI”) technologies.

The Law is not comprehensive in nature; it focuses specifically on regulating the development, deployment, and use of large (foundation) AI models in Russia.

AI definition and key objectives of the Law

The Law defines AI as a complex of technological solutions that simulates human cognitive functions and achieves results comparable to or exceeding those of human intellectual activity.

The object of regulation is large (foundation) AI models. These are software programmes that meet established criteria, with the key threshold being a parameter size exceeding 1bn.

The Law pursues three main objectives:

• providing state support for AI research and development;
• ensuring safety and technological independence; and
• enhancing competitiveness of Russian AI solutions globally.

Sovereign and national AI models

The Law distinguishes two categories of domestic large (foundation) AI models: sovereign and national, with specific requirements laid down for each. Common requirements include the developer’s Russian nationality, the localisation of data processing and computing power within domestic data centres, and compliance with Russian laws and traditional spiritual and moral values.

The Government will approve the procedure for granting sovereign or national status, the list of sectors where only such models can be used, and the model’s compliance verification framework.

Use of third-party works in content generation

Developers of sovereign and national models may use copyright and related rights-protected material for model training without the right holder’s consent in the following cases:

• if they use a legally acquired copy of the work; or
• if the copyright or related rights-protected material has been made available to the public and access to it is not restricted by technical means.

At the same time, AI model providers must notify users regarding:

• the ownership of rights to the generated content; and
• the terms for providing the user with access to the generated content, as well as the terms of its use and saving.

Labelling of AI content

The Law provides for a voluntary labelling framework for AI-generated content by users.

Information resources falling under the scope of the Law must provide users with the functionality to add the relevant labels.

The Law establishes two types of AI content labelling:

1) during content generation: owners of large (foundation) AI models must ensure that a warning about the use of AI technologies can be embedded into the generated audio, video, or image files; and

2) for posting content on information resources: owners of information resources with a daily audience exceeding 500,000 Russian users must enable them to display a warning indicating the use of AI when publishing such content.

Liability

Actors in the development, deployment, and use of large (foundation) AI models will face liability for violating the Law and any regulations adopted under it. However, the authorities have not yet introduced specific liability measures and penalties.

Secondary legislation

The Law acts as a framework; secondary regulations will supplement and specify most of its provisions. The Law outlines how authority is distributed among state bodies to adopt relevant acts and further shape AI policy.

According to a published draft resolution* of the Government, the Ministry of Digital Development will become the primary regulator in the AI sector.

Entry into force

The Law provides for the following effective dates:

• 1 September 2026: the general organisational provisions of the Law come into force .
• 1 March 2027: content labelling requirements, rules on using copyrighted works for model training, user notifications regarding rights to generated content, and obligations for developers of sovereign and national models come into force.

In addition, transitional period until 1 September 2032 is introduced. If the Government mandates the exclusive use of sovereign or national models in specific sectors, these restrictions will not apply until that date to models already used in those sectors as of 1 March 2027, provided that data is processed and stored in Russia.

Recommendations

In light of the established timelines, we advise entities subject to the Law to take the following preparatory measures:

• Owners of information resources with a daily audience exceeding 500,000 Russian users should implement functionality to enable user content labelling upon upload by 1 March 2027.
• Owners of large (foundation) AI models should ensure the following by 1 March 2027:

– the technical capacity to label generated content;
– mechanisms to notify users of rights ownership and usage terms of the generated content.

• All actors of the AI development and application sector should monitor the adoption of secondary legislation under the Law, as well as the future introduction of liability measures. We will keep you updated.

*In Russian