Kirill Chernovol, Researcher of the International Best Practices Analysis Department at the Gaidar Institute, spoke on the key innovations of the Federal law related to support of the development of the artificial intelligence technologies, which was approved on July 26, 2026. The principal provisions of the Federal law will come into force as from September 2026.
Artificial intelligence has already penetrated all areas of the economy, and regulators around the world are striving to formulate rules for its use. According to the OECD, today there are more than 80 legal acts regulating AI. In Russia, the volume of venture investments in AI in 2025 amounted to almost half of the entire venture market in the country: $70.7 of 146.4 mn, however, compared to 2024, investment in AI increased by 45.5%: from $48.6 to 70.7 mn.
Principal changes:
- The law established the legal basis for state support for Russian developers of large fundamental models, universal AI models, that run, for example, YandexGPT and GigaChat. Financial, property, and information support are provided.
- The law allowed using legally obtained and publicly available content (books, images, programs) to train national and sovereign large models. AI developers now have more certainty.
Restrictions and unsolved issues:
- The law applies exclusively to large fundamental models, which, according to IDC, make up about 17.2% of the global AI market. Highly specialized and small models remain unregulated for now.
- The law enshrines a risk-based approach to the regulation of AI, but it does not contain degrees of AI risk, general rules for their assessment and management: who should take what measures, what information about AI should be published, who is responsible for damage caused by AI, etc. In international practice (EU, Vietnam, Kazakhstan, Peru, Argentina), such classifications are already established at the regulatory level.
Read a more detailed analysis of the law in the Monitoring of legal regulation of the digital economy No.7(31).