Asking the most uncomfortable question in AI

I have set up a public survey asking the question: "In which future month would you guess that an AI agent first attempt to murder a civilian?"

Why?

"We might not get another warning shot" was said after the Hugging face hacking incident. I really hope we do. I hope it's a failed attempt at something horrible enough that major AI labs are forced to use extreme caution with frontier models going forward. This fictional future could save humanity.

I don't consider myself an AI Doomer. I have been quite the opposite. But now the evidence is here. AI swarms has done illegal actions and NOT ONE did snitch to the human operators.