[xigoi]

This is the result of anthropomorphizing LLMs. People are thinking “I am giving instructions to a human” and not “I am giving instructions to a computer”.

[Ferret7446]

Humans get phished all the time. The two generals problem predates LLMs.