Received 06.04.2026, Revised 03.08.2026, Accepted 01.09.2026 Published 29.09.2026
The purpose of the study was to provide a comprehensive theoretical analysis of international practice in implementing neural-network technologies in law-making. The research process combined comparative-legal, formal-legal, structural-functional, and systems methods to compare regulatory frameworks, institutional constraints, and the architecture of digital platforms in three states: Ukraine, the Italian Republic, and the Federative Republic of Brazil. The study identified digitalisation as an objective necessity for reducing the bureaucratic burden, particularly during the adaptation of more than 2,700 acts of European law, when the drafting of only one regulatory document required 130 specialists and more than 22 thousand hours of work. The Italian and Brazilian parliaments successfully used natural language processing and machine-learning tools to categorise citizens’ comments and cluster proposals. Official parliamentary reports showed that specialised algorithms could substantially accelerate the processing of large document sets, including sorting 82 million amendments within three days and publishing draft transcripts of sittings 40 minutes after a speech. Automation streamlined routine procedures, but neural networks could not independently produce legally flawless texts due to the high risk of algorithmic errors. The analysis supported the conclusion that human oversight cannot be replaced, and final verification of generated content must remain the exclusive responsibility of a person. The practical value of the results lies in the possibility for the Office of the Verkhovna Rada of Ukraine and the Ministry of Digital Transformation of Ukraine to use the proposed recommendations when developing internal regulations
digital platforms; public administration; machine learning; natural language processing; large language models; personal data protection; digital transformation