People who write about AI have no idea what they are talking about.
Current gen LLMs aren't magic. They are just very advanced text completion. They will do whatever they are trained on, they are not 'intelligence' and they certainly don't make 'decisions' (in the normative sense as we understand them).