Way to get hung up on a tangent. Yes LLMs are ML models, they are trained with an unsupervised learning objective, then fine tuned with supervised learning and then reinforcement learning, which are all the three main branches of machine learning. It is trained on a training set, with optimizing a training loss, with a learning algorithm. It's as ML as it gets. Do you only want to call cat vs dog image classifiers ML? Or only SVMs?

AI is a superset of ML, though today most successful AI approaches are based on ML so the line has blurred in casual speech.