The digital transformation of public administration is accompanied by the intensive development of automated information systems that process significant amounts of personal data of citizens. The practice of preparing test environments by copying records from operational databases creates serious risks of unauthorized dissemination of personal data, primarily from internal intruders with legitimate access to information resources. The article analyzes the problem of personal data protection when they are transferred to test development environments in government agencies and organizations. The analysis of the typical organizational structure and information flows is carried out, the model of the internal violator is formalized, the target threat of homicide is considered.096 (“Copying”) according to the Basic Threat Document of the FSTEC of Russia. An irreversible deconstruction algorithm is proposed that combines deterministic hashing with a unique salt based on the Keccak‑256 algorithm and context-sensitive masking. The algorithm complies with the requirements of Federal Law No. 152-FZ “On Personal Data”, Federal Law No. 187-FZ “On the Security of Critical Information Infrastructure”, GOST R57700.34–2017 and the Basic Threat Document of the FSTEC of Russia. The integrability of the solution into CI/CD pipelines is shown without reducing the flexibility of development. The results demonstrate the applicability of the proposed approach both in educational and other government organizations operating information systems with arrays of structured personal data.