You might find that strange and disagree with it with a flawed analogy but I've merely reported the official stance of the US Copyright Office and legal precedents. See, for instance, this overview with further references:

https://www.sidley.com/en/insights/newsupdates/2025/02/us-co...

"Prompts Are Generally Insufficient to Make AI Output Copyrightable"

If you don't trust this summary, read the US Copyright Office report for yourself. The gist of the position is that prompts are not specific enough and do not lead to deterministic output.

On a side note, I find it weird that even on HN people automatically assume you're only expressing a personal opinion, yet in all fairness I should have included some references from the start.