The point is, the self-policing is needed because there isn't a better policing mechanism outside of oneself for something so fundamental, so much at the frontier.
As the AI leaders themselves admit [1], they are doing what they're doing (i.e. capability-maxxing without caring deeply about actual risks this opens up for humanity) because they can and because the other guy's doing it, so why should they be left behind?
They're asking for some external force to bring the morality that they agree on paper ought to exist. There is a segment among them that are even okay with millions of people dying before the risk of AI gets taken seriously. [2]
[1] https://www.youtube.com/watch?v=wrESBnPYoZU, seek to 2:45 [2] same as above, seek to 3:25
In order to change the game theoretic outcomes, we‘d need a systemic change that affects the rewards, not a personal attitude change that will become a losing strategy in the game.
Also, do you remember how tobacco companies were invited to the table to discuss whether smoking is bad for you? Were those the days of personal responsibility or was it even before that?