Claude (and LLMs in general) takes an input and generates an output as a general purpose tool.
A few years ago I would have handled a request from upper management to my department that would need new software to fulfill by writing up such a description and then assigning it to one of the programmers in my department to implement.
I certainly would not then go around saying I built that solution.
Nowadays it is quite possible that in that situation I'd give that exact same written request to an LLM.
I don't see how I could now reasonably say that I did build the solution, since the only difference is in who/what I gave my problem description to.