A recent analysis investigates the knowledge cutoffs of major AI models like Claude and GPT to uncover details about their pre-training timelines. By evaluating the specific information these language models retain, the study provides unique insights into how leading developers curate and finalize their massive training datasets.