I am still doing the research, but my feeling is that LLMs play tit for tat - if you don't provoke them, they are told to cooperate

Here are some initial results -

https://thenunchiproject.com/blog/ai-betrays-you-in-the-last...