Expert reviews Russia’s proposed framework for regulating artificial intelligence
Russia’s Mintsifry (Ministry of Digital Development) has published a draft law to regulate artificial intelligence. The document defines key terms, outlines the rights and responsibilities of market participants, introduces content labeling rules, and states that sovereign, national, and trusted AI systems must be developed within Russia.
In comments to IT Russia, Dmitry Sytsko, chief information officer at BCS Bank, said the law is built on a sound logic.
“Regulation is based on a risk approach, much like in the European AI Act. The key difference is that Europe introduced detailed mandatory requirements across a broad range of systems, while Russia opted for a more flexible model: the commercial sector is exempt from binding rules, with strict requirements applied only where risks are genuinely high,” he said.
The law is expected to take effect on September 1, 2027, giving the industry time to prepare. Its framework nature is both a strength and a sign that further development is needed.
Balance and Public Consultation
“The commercial sector is exempt from mandatory requirements, labeling of AI-generated content is finally being formalized, and copyright for outputs produced by neural networks is assigned to the user if they make a creative contribution. All of this is reasonable and long overdue. However, key decisions – such as criteria for ‘trusted’ models, conditions for mandatory certification by FSTEK and FSB, and rules for maintaining a registry – are left to future government regulations that have yet to be drafted. For now, businesses and developers are given a framework without detailed implementation,” Sytsko said.
According to the expert, the concept of “sovereign” and “national” models requires careful consideration.
“The idea of developing models exclusively using Russian datasets, by Russian citizens and companies, without foreign components, is understandable as a path toward technological independence. But global practice shows that the fastest way to build a strong model is to train it on top of an existing one. What matters here is balance, and public discussion is exactly the stage where the industry can offer workable formulations to regulators,” Sytsko said.
The draft law is currently under public review.
