The article focuses on the implementation of advanced language models (LLM) in insurance companies and the development of a comprehensive approach to their safe and effective integration. It explores architectural solutions that ensure the verifiability and transparency of model operations, methods for evaluating cost-effectiveness (TCO/ROI) and risks, regulatory requirements under the EU AI Act and NIST AI RMF, and the importance of long-term risk management associated with the implementation of generative AI. The article emphasizes the role of human oversight, independent validation, and the development of internal corporate standards that ensure the fairness, reliability, and sustainability of AI-based solutions. The proposed FRIA-in-the-Loop and LLM-Assurance for Insurance concepts serve as a foundation for creating a comprehensive quality assurance system for the use of generative technologies in the insurance industry.