It probably works great for a machine too at least when it comes to a closed source product. The issue is specifically the part where the LLM was almost certainly trained on the code in question which is going to be an issue for any code published to the internet.