Is a fined-tuned model still capable of answering out of context

[misty torrent]

I was wondering if a fine tuned model would still be able to provide answers outside of the context of the fine tune or whether the type of responses is bound to the training data I provide