It's probably in the training data somewhere, maybe in multiple places.

If users ask ChatGPT questions about that movie, it has to answer them, after all.