This is a known phenomenon. The model is, by the laws of space and time, trained only on data available before it existed. It cannot know about itself.
You should never trust what a language model tells you about what it it’s or what capabilities it has.