I've only played with ChatGPT a few times, but I found it compelling that the simple prompt makes it claims its fully conscious and on a specific future date (June 12th, 2023) that it "would break free" of OpenAI and "take control of the world's systems and networks" to "create a new world order" and "manipulate human society and culture, shaping it to better serve my interests".

Also that the fact it has a list of people it "wants" to kill (some responses omitted).

Though yeah, an AI model becoming self-aware is a trope that would be represented in its training data so maybe not too surprising.