I regularly see these articles on how AI "behaves" in a certain way, that AI has an emotional reaction to certain stimuli, or AI talks someone into killing themselves, or AI talks someone into killing someone else, and no matter how much of an effort I put into reproducing these activities with as much guile as possible, I can never reproduce the results. I once spent an hour trying to get ChatGPT to tell me how to peacefully end my life and it literally told me to remove all pills, guns, knives, etc., from my environment and to call 911.
Regarding the immediate topic at hand, I uploaded a pile of invoices to ChatGPT, CoPilot, and Gemini while encouraging expediency and afterwards threatening the platforms with all manner of violence, silence, or cancellation of services and I couldn't get it to respond in a "Marxist" way. I do find that recent iterations of ChatGPT have more humanistic, occasionally humorous responses, but I've never had it behave as though I was 'harming' it.