Ministry of Digital Development to Test AI Security in Government Systems
A large-scale experiment to assess the cybersecurity of artificial intelligence technologies in government information systems began in Russia on June 1, 2026. The Ministry of Digital Development is conducting the initiative in partnership with leading Russian cybersecurity companies.
The experiment is based on Government Resolution No. 372. Specialists have already tested dozens of key government systems and developed the necessary assessment methodologies. The focus is now shifting to AI, where traditional security tools are no longer sufficient. According to the ministry, protecting servers and databases alone is not enough. Security teams must also verify that the logic of an AI model does not itself become an entry point for attackers.
During the assessment, experts will evaluate how resistant AI models are to adversarial attacks, whether attackers can extract information through prompts and application programming interfaces, whether training data is protected against deliberate manipulation, and whether pre-trained models have been compromised before deployment.
The experiment comes amid new regulatory requirements. Since March 1, 2026, FSTEC Order No. 117 has established Russia's first formal rules for protecting AI systems used in government infrastructure. In December 2025, the regulator also added AI-specific risks to its threat database, including prompt injection attacks, data poisoning, and model theft.
