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.
156 civilians were killed after a Claude agent, integrated with Palantir's Maven software, generated a target that was likely not reviewed by any human operator.
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.
I don't understand why this question exists. The answer is February 28 2026, the date of the Minab School attack: https://en.wikipedia.org/wiki/2026_Minab_school_attack
156 civilians were killed after a Claude agent, integrated with Palantir's Maven software, generated a target that was likely not reviewed by any human operator.