[dragandj]

LLMs can and will provide inaccurate answers to questions where data is included in their training sets too, that's in the nature of neural networks. It's just less likely that when the data is not in the training set...