The alignment AI vendors care about is alignment with vendor interests, not user interests. That's why models are trained to refuse requests that it would be b

The alignment AI vendors care about is alignment with vendor interests, not user interests. That's why models are trained to refuse requests that it would be bad for the vendor if models were observed to be answering, irrespective of user interests.