If AI really is dangerous then this is an argument for all AI research to be required to be done fully in public view. No proprietary models, no closed platform plays, nothing that isn't done fully in the open. Sunlight is the best disinfectant. That way if something starts to behave in really dangerous alarming ways we can all learn about it and prepare to deal with it.
This was OpenAI's original mission before Altman took it private and closed. Remember?
The most plausible dangerous scenario for AI is that it will end up in the hands of a small number of people who will use it as a force multiplier to control and conquer other groups of people at tremendous scale. Other scenarios like autonomous AGIs trying to take over or exterminate humans are significantly less plausible, especially since these models so far have precisely zero ability to self-reflect or exercise independent agency. They only do what we tell them to do after being trained on data that we created and fed them.
Now check out that "worldcoin" thing and ask me if the people behind that (and behind appropriating and closing a non-profit) are the people who should be in exclusive control of closed massively powerful AIs with a regulatory moat in place to prevent anyone else from creating anything that equals them.
The most plausible of the scary scenarios is the one Altman is trying to bring into being: a small oligopoly empowered with AI beyond anything anyone else possesses. He's either completely delusional or a grifter... or perhaps both since the best grifters are often delusional.
Some of the shadiest people I have ever met are full of hot air about how enlightened they are and the purity of their motives. When someone starts tooting their own horn about how pure and wise and altruistic they are, check their basement for mounds of loose dirt and their closets and drawers for little baggies of human hair and clippings of clothing.