ELISAR represents a significant advancement in cybersecurity risk assessment through its novel modular architecture, which integrates Zero-shot Chain-of-Thought (CoT) prompting to train a local Large Language Model (LLM). This LLM not only excels in complex reasoning tasks but also capitalizes on the historical knowledge of past risk assessments stored in a secure, local environment, ensuring data privacy and confidentiality. By leveraging a Retrieval Augmented Generation (RAG) system, ELISAR enhances the precision of semantic data extraction from a vector database, improving decision-making for cybersecurity professionals. The combination of state-of-the-art Generative AI (GenAI), LLMs, and a focus on data confidentiality allows for the development of AI-driven assistants that preserve sensitive information while delivering superior performance in risk assessment use cases. Compared to traditional GenAI models like GPTs or AI copilots, ELISAR offers a more secure and effective solution tailored to the unique challenges of the cybersecurity field.