“Laredon has developed a customized Llama-based model designed to act as a guardrail, preventing employees from submitting prompts that are illegal or violate company policy.”