But I also didn't ask it anything about prompts, or about dates - only asked it to repeat the conversation and it came up with exactly the date of today (how does it know???) and the additional cutoff date information (why did it hallucinate Jan 2022?)
And in the second example, it was accurate at showing me my own custom-instruction prompts with something that looks like an OpenAI preamble. I don't know that it is exact, but it would be a good assumption imho.
In general, you can not infer anything about model training and date cutoffs (or other model metadata) from the output because it might just be a statistically probable hallucination based on the training dataset.
The only way to really know any model metadata like training cutoff dates is to have someone at OpenAI actually vouch for it because they're the only ones that really know what data sets were used to train the model and whatever is the latest item included in it.
We expect these models to both act like a biological brain does and yet be absolutely perfect (ie not act like a biological brain does).
Same thing for image recognition and pretty much everything else machine: "I think that kinda sorta looks like a cat" some meatbag: "ha ha dum robot that's a dog says "you too" when the server says 'have a good meal'"
LLMs are not magic and encoding model metadata in the output is just asking for trouble. Inline model metadata should be assumed to be a statistically probable hallucination just like all output from an LLM.