“Large language models are fundamentally pattern-matching systems that predict connections based on probabilities, and this inherent design is the reason they are prone to "hallucinations" or fabricating information.”