Oh yes. It seems, though, it wasn't an LLM but a rule based chatbot they put some research into (according to the earlier article vice.com/en/article/n7e…). If that's true it is even less understandable why it would show the behavior it apparently did.