I think it's really instructive that only closed-model companies are doing these kind of "my AI could kill the world" demonstrations. All while telling us how dangerous open models are.