Who's accountable when AI writes the code?

I use AI to build software. It helps me get ideas working faster, handle repetitive tasks, and explore different approaches without spending days on each one. As a developer, that's useful. As a client, you should benefit too: you get working software sooner, and you pay for less routine work.

But generating code doesn't mean the job is done. Someone still needs to understand your business, decide what to build, and check whether the result actually solves the problem.

A feature can look finished while still handling payments incorrectly or exposing data it shouldn't. Recently, AI generated an endpoint for me that checked whether someone was logged in, but not whether the data they asked for belonged to them. It worked perfectly in normal use. Changing one ID in the request, though, would have exposed another customer's information. I caught it during review, before it reached anyone.

That's exactly the kind of thing someone needs to check before it reaches your customers, and that responsibility sits with me. If I deliver software to you, I'm accountable for it. "The AI wrote that part" would be no answer at all to a client whose website has stopped working, or whose customers' data has leaked.

So every change gets reviewed by me before it ships. I need to understand what's going into the product, how it fits together, and how I'm going to maintain it.

I also need to explain my decisions. Why this integration? Why a custom solution? Why does something that looks simple need more work? These choices affect what you pay today and what the software costs to run later. You deserve clear answers.

AI makes writing code faster. That means I can spend more of my time checking assumptions, testing the result, and thinking through what happens after launch. That's where a lot of a developer's value sits: judgement, understanding the business, and being someone you can call when something goes wrong.

The tools will keep changing. My name is still on the work.