Because in the end, you, the person that is forced to use various AI chatbot/agent/model/whatever, will be held responsible for anything that happens after one of the 3000 decisions they imposed you to make with no way to check everything turns out to cause the slightest problem. When that happens, YOU were supposed to know that NOTHING the AI tells/says/do is to be expected correct, so it's your responsibility if something's gone wrong.