The Cabinet of Ministers of Ukraine has adopted a resolution that launches the creation of a national data corpus for training and testing large language models (LLMs). It will be overseen by the Ministry of Digital Transformation.
The corresponding document No. 1416 was approved by the government on November 11, 2026. The Ministry of Digital Transformation was designated as the curator and organizer of the database, while the state enterprise "DIIA" will be responsible for its technical content, structuring, and protection. The Ministry of Digital Transformation has three months to deploy the corpus. After that, the ministry, together with the Ministry of Education and Science and the National Academy of Sciences, must develop proposals within six months regarding the further development of the national system of language resources.
The corpus will consist of anonymized structured and unstructured texts in Ukrainian: databases, literary works, scientific research and
Executive authorities are obliged to transfer their available open datasets within a month after a request from the Ministry of Digital Transformation or the state enterprise "DIIA". Private companies, libraries, educational institutions, and individuals may join the filling of the database voluntarily on the basis of license agreements in compliance with copyright.
Access to the formed data corpus will be provided free of charge under a license agreement, but not everyone will be able to obtain it. Only Ukrainian legal entities and sole proprietors who have confirmed experience in developing or testing LLMs and undertake to use the data exclusively for this purpose have the right to apply to work with the database. At the same time, the resolution establishes strict restrictions: access is closed to companies from sanctions lists, financial institutions, investment funds, as well as businesses operating in temporarily occupied territories, in russia, belarus, or having ultimate beneficiaries there or shareholders with a stake of 10% or more.
Applications for access to the "DIIA" State Enterprise will be reviewed within 10 working days. If misuse of data or violation of security requirements is detected, access to the building will be terminated immediately. The document also provides for an intellectual property protection mechanism: if a rights holder complains about the illegal use of their texts, access to the disputed dataset will be temporarily blocked within one working day pending verification.