Russian government approves AI bill with business-backed amendments

The government commission on digital development has approved a bill on the development of neural networks in Russia. Before the vote, amendments were made to the document to reflect feedback from the business community. For the first time, the tech industry's position has officially made it into the text.

Why this law is needed

Russia still has no dedicated legislative act regulating the creation and use of AI systems. Neural network developers operate in a hybrid legal environment: general rules on data protection, on liability for service quality and on IT security apply, but there is no special regime for AI.

This creates problems on both sides. Businesses do not know exactly what kind of oversight awaits AI products, so they play it safe. Regulators lack the tools to intervene when an AI system causes real harm.

The bill is meant to close that gap: it introduces basic definitions, identifies who bears responsibility and creates a mechanism of regulatory "sandboxes" for testing solutions before the formal regulatory framework is adopted.

The debate over whether a special law is needed has run in Russia for several years. The current acceleration is tied to the fact that AI systems are increasingly used in public services, healthcare and the financial sector, and the legal vacuum is no longer an abstract problem.

How business shaped the text

The first draft triggered a wave of objections from technology companies and industry associations. The complaints came down to a few recurring themes: vague wording that leaves too much room for interpretation; requirements that duplicate existing personal data legislation; and an excessive burden on developers of systems that pose no obvious risks to society.

Business representatives pushed for a more differentiated approach: strict control for high-risk uses and lighter rules for everything else. Given that the bill was revised before being submitted to the State Duma, some of these positions appear to have made it into the text.

• Dialogue took place through working groups at the relevant agencies
• Amendments were made at the government approval stage, before the first reading
• Several industry associations publicly backed the updated text
• Particular emphasis was placed on removing provisions that duplicate existing laws

What happens next

Approval by the government commission is an intermediate step. The next one is the first reading in the State Duma, where the document will again be open to amendments. Technically complex laws in Russia usually go through three iterations, and each one risks making the text both better and worse.

The international context adds pressure: the European AI Act has already entered its transition period, and companies operating in several jurisdictions have to reconcile Russian rules with European ones. The wider the gap between requirements, the higher the compliance costs for business.

Even after the law is adopted, the work will continue: most provisions will require secondary legislation, such as methodological recommendations, standards and registries of AI systems. That will take another one to two years.

"It is important not just to pass a law, but to pass a workable one,

which creates clear rules rather than a new field of uncertainty," industry representatives sum up.

What this means

The fact that business feedback was taken into account before the first reading is a signal in itself. Russian technology regulation has historically reacted to problems that had already emerged rather than consulting preventively. If this approach holds in the final text of the law, the industry will get a more predictable environment, which for AI development matters more in the long run than any subsidies or tax breaks.