Use of large language models is transforming risk assessment

The use of large language models (LLMs) is transforming risk assessment, customer engagement, underwriting and policy personalization today and their creators continue to release multiple advances. In the past quarter, Reinsurance Group of America (RGA) researchers say advances in LLM technology have been marked by significant improvements in dynamic reasoning and task-specific adaptability.

These findings and others are summed up in a new article from the reinsurer, entitled GenAI in Insurance Update: Q2 2025. In it, RGA names and discusses the different LLM innovations, including those out of China, and discusses breakthroughs in quantum computing which promises to enhance risk assessment and fraud detection “potentially uncovering patterns invisible to classical computers,” they write. Recent quantum computing breakthroughs by Microsoft and Google potentially pave the way for scalable quantum systems, they add.

Quantum computing 

“These advances push quantum computing closer to real-world applications and signal a future that revolutionizes complex problem-solving, from materials science to artificial intelligence. For the life insurance industry, quantum computing presents game-changing potential in risk assessment, portfolio optimization and fraud detection. Quantum algorithms could vastly improve the accuracy of actuarial models,” the note states. 

Coming back to a discussion about the technologies actually available for use today, the report also looks at open-source LLM models and closed models and the benefits and drawbacks of each. In short, they say open-source models provide transparency, flexibility and cost advantages, while closed models include robust safety guardrails and higher reliability in maintaining compliance with evolving regulation. “These models are optimized for accuracy and require less in-house expertise to manage,” they write. “Insurers must carefully weigh their considerations when selecting AI solutions, balancing the need for customization with the operational risks associated with model governance.”