You're overthinking it. You're building on your own definition of what a tool is, but courts are likely to find that a person who used an AI with a specific type of prompt were using it as a tool and are responsible for the clearly stated intent behind their use.
Whether that's actually legal in this case I don't know, but I'm pretty sure courts won't conclude "welp, it was the AI, not the user" in a case like this.
Yeah, probably you're right. Though it is an interesting thing to overthink - we have decided that LLMs are not tools for copyright but are tools for accountability. At some point they'll get smart enough to be accountable, and what happens then?
Whether that's actually legal in this case I don't know, but I'm pretty sure courts won't conclude "welp, it was the AI, not the user" in a case like this.