Proceedings of the International scientific and practical conference ―Science at the Turning Point of History‖ (May 25-27, 2026) / Publisher website: www.naukainfo.com. – Lviv, Ukraine, 2026. - 362 p.

165 RestSharp, or a database snapshot - is forwarded to the LLM agent, it must undergo automated scrubbing. Using deterministic regular expressions or lightweight named- entity recognition (NER) models, the framework programmatically replaces sensitive entities (such as credit card numbers, production tokens, and real user emails) with synthetic tokens. This ensures that even if a model retains parts of the input text during inference, the stored data contains no actionable or compromising information [3]. Furthermore, governing AI interactions within closed automation loops necessitates strict compliance with Role-Based Access Control (RBAC) and comprehensive audit logging. Enterprise QA environments must trace every request made to the internal LLM, logging which test suite triggered the call, the token consumption, and the model's output. By applying data retention policies that instantly purge inference contexts from the model‘s cache after the test execution session concludes, organizations eliminate the risk of accidental internal data exposure. Implementing these combined security practices enables enterprises to build resilient, AI-augmented quality assurance ecosystems that maximize test maintenance efficiency while maintaining an uncompromised posture toward data privacy and regulatory compliance. REFERENCES: 1. Appium. Official website. URL: https://appium.io/docs/en/latest/ 2. National Institute of Standards and Technology. (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0) (NIST AI 100-1). U.S. Department of Commerce. DOI: https://doi.org/10.6028/NIST.AI.100-1 3. OWASP Foundation. (2025). OWASP Top 10 for Large Language Model Applications. URL: https://owasp.org/www-project-top-10-for-large-language- model-applications/