Abstract: The object of research is neural networks, the VKontakte platform, the Telegram messenger, the Python programming language and its libraries, and a block diagram of a computer system model. The subject of the study is a computer technology for detecting extremist content in text form and specific groups containing it on the VKontakte social network. The authors consider in detail the structural scheme of the computer system model, the functional modules included in it, and illustrate their interaction. The paper uses a pre-trained model designed for processing the Russian language, and provides conditions for ensuring high accuracy of recognition of illegal content without signs of retraining. The paper presents the results of checking the test data confirming the operability of the computer system. The proposed prototype of the computer system ensures its integration with the Telegram messenger, which increases usability and facilitates the process of generating queries and reports. The novelty of the research lies in the creation of a prototype of a computer system for searching and detecting extremist messages on the VKontakte social network using the Python programming language and the VKontakte API programming interface (VK API). The basis of the prototype computer system is a neural network that works with the Тгаnsformers and Тоrch. A special feature of the computer system is the ability to analyze messages on a social network and subject them to binary classification for the content or non-content of illegal information in messages. The main conclusions of the study show the efficiency of the system, the simplicity and convenience of its use, the possibility of detecting illegal text content. A distinctive feature of the prototype is the ability to detect illegal content presented using slang expressions.