I spent some years as a product manager. This might be a cynical take, but my relationship with AI-written code is largely the same as it was with human-written code.

I would research and explain what I wanted the product to do. We'd discuss technical trade-offs, drawbacks, and risks. Eventually, we'd ship something. I understood broadly how the infrastructure worked, but to a large extent I had to trust the quality of the code my colleagues were writing. Sometimes things broke. That's okay, we're only human.

I think I'm just more comfortable being held responsible for something I don't fully understand because I always have been. Maybe I'm delusional and the comparison won't hold. We'll see.