You need to know: the work of the checkpoints of goods will be unified using neural networks

Russia is developing a new generation of AI models to automate data on all goods exported from the Russian Federation and imported into the country. Scientists from the REC of the Federal Tax Service of Russia and the Bauman Moscow State Technical University plan to increase the efficiency of commodity flow analysis and pricing in international trade by 70%. To achieve this goal, specialists are working on the unification and automation of data processing coming from the checkpoints of goods in Russia, which are now performed manually. How the new technology will work and to whom it will be useful — in the exclusive Izvestia article.

An AI model for automating product data

The Scientific and Educational Center of the Federal Tax Service of the Federal Tax Service and the Bauman Moscow State Technical University are developing models for automated analysis of data on imported and exported goods. Machine learning and AI algorithms are expected to increase the effectiveness of commodity flow analysis and pricing in international trade by 70%. The model has already demonstrated high accuracy in extracting key characteristics of product items and delivery conditions. For example, the system successfully identifies trade names, quality standards, geographical logistics parameters, and other significant attributes to assess market conditions.

"For one of the product lines, the system correctly identified the product brand for 98% of the dataset lines and the exact delivery point for 99.6% of the lines," said Alexey Borodulin, director of the REC of the Federal Tax Service and the Moscow State Technical University.

The development will significantly increase the effectiveness of analytical work in regulatory authorities, including in monitoring cross-border transactions and detecting price anomalies. The introduction of AI tools will also help to redistribute labor costs: automation of routine operations will give experts the opportunity to focus on deeper data interpretation and risk modeling.

— Currently, the identification of goods in customs declarations is carried out manually by specialists of the Federal Customs Service of Russia, which complicates the detailed analysis of large amounts of data. For example, certain grades of oil can be supplied through certain checkpoints. The same checkpoint may have different spelling — such a difference may depend both on the source system (for example, on the format or requirements of a specific information base), and on accidental or intentional incorrect information, - Alexey Borodulin added.

Mind Games: Most Russian studios are introducing AI into the development process

What tasks do they solve using neural networks?

Therefore, one of the main tasks of the new technology is the unified analysis of data on product items as they pass through checkpoints. This will improve the correctness and reliability of import and export controls. In addition, scientists have automated the classification and extraction of product characteristics using modern data mining technologies.

According to him, it is already planned to introduce the development into the work of the Federal Tax Service for automated monitoring of actual and dynamic price indicators for exported goods. This process is scheduled for 2026 and will be carried out based on the results of the piloting of the project.

— Based on the general information, the model may be of interest to the Federal Customs Service. We are ready to evaluate the prospects for its use in customs control of goods in the event that the Federal Customs Service of Russia receives presentation materials and algorithms for a specific model for automated data analysis, the press service of the department told Izvestia.

The editorial board also sent a request to the Federal Tax Service of Russia.

From customs declarations to logistics

If scaled to new product groups, the development looks promising for implementation in the work of the Federal Tax Service. This technology will be useful for the state in terms of more precise control over cargo flows, pricing, as well as regulation of tariffs, customs duties and excise taxes, Artem Kiryanov, deputy chairman of the State Duma Committee on Economic Policy, told Izvestia.

The advantages of development are automation of routine processes, which allows experts to focus on strategic tasks such as risk modeling and deep data interpretation. Anton Averyanov, CEO of the ST IT Group of companies, TechNet NTI market expert, noted that similar systems in other countries have already demonstrated a two- to three-fold reduction in time for processing declarations. He sees similar potential in Russian development.

— The model successfully manages to identify attributes that are applicable for assessing market conditions, which will help in monitoring anomalies, and this is already important for determining possible dumping or, worse, artificially lowering prices for tax evasion. In the long term, this may lead to the creation of a single platform for real—time analysis, where data from various sources — from customs declarations to logistics — will be aggregated for a comprehensive assessment," the expert said.

According to him, when implementing such projects using AI, it is important to take into account the need to ensure cybersecurity and compliance with personal data protection standards in accordance with federal law. To maximize returns, it is necessary to invest in training Federal Tax Service personnel so that they can effectively use neural networks for strategic planning, and not just for routine monitoring.