Federal Antimonopoly Service

Rospotrebnadzor and FAS to Extend Oversight to Websites: Automating Foreign Word Replacement Using LLM

Russian compliance experts have developed a specialized microservice called Normograph that combines multi-stage filtering, OCR, and LLM processing to help organizations meet the requirements of Federal Law 168-FZ on protecting the Russian language. The system automatically identifies prohibited foreign borrowings and Latin script on websites, cross-references them against official dictionaries approved by the Russian Academy of Sciences, and suggests context-aware Russian replacements while preserving marketing meaning and grammatical agreement. It excludes registered trademarks, brand names, and terms without Russian equivalents using dynamic white lists and Rospatent data. The solution processes pages up to 20 times faster than manual review by filtering out already-compliant words before sending only problematic fragments to the language model. An OCR module based on Yandex Cloud Vision extends checks to images and banners. The service was built with GigaChat API but remains provider-agnostic and avoids sending full pages or confidential data to foreign AI services.