As of September 1st, provisions of the artificial intelligence law have come into effect in Russia: the document for the first time defines AI in legislation and outlines measures to support developers.
The law introduces rules for large foundational models — systems with at least 1 billion parameters, capable of performing complex intellectual tasks at or above human level. The concepts of sovereign and national AI models are introduced: sovereign models must be entirely developed by Russian companies, while national models may use foreign components and open-source solutions. The government will determine measures of state support for developers of such models, as well as cases where organizations may only use sovereign or national AI solutions.
Separate provisions of the law will come into effect in March 2027: content labeling created with AI, and requirements for using intellectual property results to train such systems. For developers, legal certainty is introduced: sovereign models will be eligible for state support and priority in the public sector, while national models will face more lenient requirements. For users, mandatory labeling of generated content will be introduced from 2027, which will reduce the risks of disinformation.
At the Eastern Economic Forum, experts discussed how to motivate employees to use AI: without salary increases, the benefits of technology for workers are not obvious. Alexander Sirovatksy, Managing Partner of Media Instinct Prime, stated that the company is testing a four-day work week as an incentive and sees potential for 7% annual productivity growth. According to him, this format motivates people to master new tools faster.
"Pervyy Tekhnicheskiy" reported that Russia is developing a package of tax incentives for creators of national and sovereign large foundational artificial intelligence models.