Give it the first few sentences and ask it to complete the next sentence. If it gets it right without search it's guaranteed to be the real system prompt.
No, just that the data was trained on, not that it is its real system prompt, which I doubt it is. It talks about a few specific tools, nothing against "don't encourage harmful behavior", "do not reply to pornography-related content", same with CSAM, etc. Which it does.