I do not think a ban would be efficient. It would just force those who use AI to dissimulate it. I see it a bit as drug use, banning hasn’t solved the problem ; and the countries who handled it the best were those regulating the use.
So I think we should look for a good regulation. It is more complicated to set up, but it has a chance to work.
In my mind, it should allow some projects to identify as not using any AI, and request that each project define what is their use of AI. Ideally, we could know this without opening the project (maybe thanks to a badge).
Also, shaming the use of AI only has the same aim as banning it: forcing it to grow hiden and out of control.
I would say that the good answer is “mox” (not sure of its english name, but it is likely similar. Used uranium is extremly dangerous for a very long time (millions of years). Yet it could be recycled and used again, and the result would be extremly dangerous for about 500 years (which is still far too much, but less crazily impossible to manage).
We already have enough used uranium to cover the next 200 years of needs, at least… And so it would be a lesser harm for future generation if we use mox than new uranium…
(But the true answer is we need new bombs, so new uranium…)