The rise of generative AI has increased awareness of underlying technologies like large language models (LLMs), which are neural networks trained on large volumes of text, images, and video to recognize patterns and generate natural language. For mission-critical use cases, LLMs must be tailored for a specific purpose, as risk tolerance in the mission space differs from casual ChatGPT use. LLMs like ChatGPT have billions of parameters; shrinking them can reduce resource use, extend access to the edge, and yield more accurate responses. GDIT recommends asking five questions when using AI and LLMs for the mission: 1. Will I maintain control of my data, including in-house emails, records, file shares, and document repositories, to enforce governance, privacy, and compliance policies? 2. Can I securely share that data with the LLM, ensuring it is protected at every stage as the dataset grows? 3. What does continuous improvement look like, including how the LLM takes in new information and persists changes for sustained accuracy? 4. How well does the LLM understand the mission space, including what data it was initially trained on and how it performed? 5. Is my model protected against data poisoning?