I dont use any of these tools but a yearish ago, my junior was trying to cheese his way thru his bachelors project by getting chatgpt to write the code and it was plain garbage. You would ask it to correct one thing, only for it to undo the previous correction. All of this only possible, mind you, if you can translate the code to the math it is doing. How you'd develop this skill if you don't write things for yourselves is beyond me. I would rather take badly written Fortran over generated python or whatever every day.
Here's another worry: who do i hold accountable with generated code? When I use X well established software, or use code written by Y well known person, there is a guarantee that the work isn't filled with a buttload of bullshit because of wrong code. But when the code is generated by a machine, I just dont trust it. I don't think it is easy or possible to test it even by a so-called expert given the experience (numerical prog is a devil)