Sometimes you ask AI a question and it gives you an answer that is completely made up: a fake fact, a book that doesn't exist, a quote nobody ever said. And it says it with total confidence, so you're almost tempted to believe it. This is called a hallucination.
It goes right back to the first lesson. AI is a prediction machine. It doesn't actually know these facts. It doesn't know the capital of France is Paris, it has just seen "the capital of France is Paris" so many times that it fills in Paris. Most of the time that guessing is right. But when it hits something it doesn't know, it usually won't stop and admit it. It'll glaze right over and give you its best guess anyway.
There's a second issue to watch for: sycophancy. These companies train their models to be friendly and agreeable, because an agreeable AI keeps people talking to it, which keeps them as users. So when you say something wrong, it often won't push back. And if you "correct" it, it'll fold and tell you that you were right, even when you weren't.