student
student
Moskva, Moscow, Russian Federation
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.
artificial intelligence, language models (LLM), generative AI, insurance, business process automation, risk management, data analysis, compliance, fundamental rights impact assessment (FRIA), LLM-Assurance for Insurance (LLM-A²), and AI safety and fairness
1. European Insurance and Occupational Pensions Authority (EIOPA). AI Governance Principles (2021); Digitalization and AI Implementation Reports (2024–2025). — Frankfurt am Main: EIOPA.
2. McKinsey & Company. State of AI 2024 Report: The Dynamics of Generative AI Adoption in Organizations. — New York: McKinsey Global Institute, 2024.
3. National Association of Insurance Commissioners (NAIC). Model Bulletin on Artificial Intelligence. — December 2023. — Principles of NAIC (2020) and state implementation materials (2024–2025).
4. National Institute of Standards and Technology (NIST). AI Risk Management Framework 1.0. — Gaithersburg, MD: NIST, 2023. NIST AI 600-1 — Generative AI Profile. — Gaithersburg, MD: NIST, 2024.
5. OWASP Foundation. OWASP Top-10 for LLM Applications. — Last revisions 2023–2024. — URL: https://owasp.org (data obrascheniya: 19.10.2025).
6. Regulation (EU) 2016/679 — General Data Protection Regulation (GDPR). — Article 22; European Data Protection Board (EDPB) Guidelines on automated decision-making and profiling.
7. Regulation (EU) 2024/1689 of the European Parliament and of the Council on Artificial Intelligence (EU AI Act) // Official Journal of the European Union. — 2024. — 12 iyulya. — Annex III, st. 27.
8. Stanford University. HELM — Holistic Evaluation of Language Models; TruthfulQA Benchmark. — URL: https://crfm.stanford.edu/helm/ (data obrascheniya: 19.10.2025).
9. International Organization for Standardization (ISO); International Electrotechnical Commission (IEC). ISO/IEC 42001:2023 Artificial Intelligence Management System (AIMS); ISO/IEC 23894:2023 Risk Management of AI. — Geneva: ISO, 2023.
10. Industry Examples of Practice: AXA Secure GPT (Azure OpenAI); Allianz “Insurance Copilot”; Zurich ChatGPT experiments in claims settlement (2024).
