Russia has opened a draft law on artificial intelligence for public discussion. The bill is intended to set rules for the development and use of AI, and if adopted it could take effect on September 1, 2027. The Digital Development Ministry said the new rules are meant to protect citizens from hidden manipulation and discriminatory algorithms.
The draft divides AI into several categories, including “sovereign,” “national,” and “trusted” systems. For tools used in the public sector and critical infrastructure, it proposes a separate registry and review by the FSB, according to Kommersant. The bill also formalizes a policy of technological independence: the development, training, and use of such models would have to take place inside Russia, relying on domestic infrastructure and specialists.
The document separately lists regulatory principles, including security and adherence to “traditional spiritual and moral values.” Companies would have to inform users when goods or services are provided with AI and no human participation, and also when an algorithm makes a decision affecting a person’s rights. In some cases, citizens would be able to refuse that mechanism and request a standard non-AI way of receiving the service.
The bill gives people the right to challenge decisions made with the help of algorithms, including through a pretrial complaint procedure, and to seek compensation for harm caused. Developers would be required to reduce discrimination risks, assess threats in advance, and document how systems work. Operators would have to test AI regularly, track incidents, and suspend operation if there is a risk of harm.
It also proposes mandatory labeling of any content created by AI. Platforms with more than 100,000 daily users would have to monitor whether such labels are present and, if not, add them or remove the material. The document also provides support for data centers and supercomputers through incentives and state funding.
“How the agency would check how patriotic or merciful a neural network is is a very difficult question,” said Yuri Borisov, a partner at Digital & Analogue Partners.
Experts say the draft combines European, American, and Chinese approaches to AI regulation. Alexander Tyulkanov of AFNOR’s AI standardization commission told Novaya Gazeta Europe that stronger state involvement distinguishes the Russian model from the European one, where requirements are more often set through technical standards and independent assessment.