Knowledge cutoff
Definition: The knowledge cutoff is the point beyond which a model has seen no training data: it is unaware of later events.
That is why a model can be wrong about recent news without access to an external source. Web search or RAG fills the gap.