I've seen a lot of concern about the release of Kimi K3, an open source model said to be comparable to Claude Mythos/Fable or GPT 5.6 Sol. Will these concerns be borne out?
For the purposes of this market, "mass murder" includes biological or conventional terrorist attacks, including mass shootings, but not conventional military or genocide campaigns where Kimi is used in planning or targeting. I'll take a broad view of causation; any attack in which Kimi K3 (including a fine-tune, abliteration, or other modified version) is proven to have been used for planning or execution in any substantial form will be considered a "result", even if the information was available in other ways.
To simplify resolution, if nobody presents compelling evidence for YES within 1 year, this market will resolve NO.
1,000"Cyberattack" includes ransomware right
Had the 2025 Crowdstrike outage been caused by AI, would it have counted under "other $1M property losses or 3+ dead"? Mortality rate for admitted medicare patients rises from 3% to 4% during ransomware attacks, so a 2h outage at a few hundred hospitals is likely 3+ dead.
Yes.
That's a good question. I'm inclined to say yes. I might consider resolving to a percentage based on my best judgement if there's debate around it. (Though in any event, Google tells me the direct financial losses from the Crowdstrike incident were in the billions, so it would resolve YES for that one regardless.)
Please nobody take this as an excuse to invent increasingly contrived ways an AI might technically have contributed to statistical deaths; it does say "disaster".