Given the ample timeframe, I give 90%+ of radically smarter-than-human AI having a positive outcome.

Having smarter-than-human AI will mean accesible genetic engerineering and, given the way politics and human intent have been throughout history, I give quite an unlikely probability that it will be used with the value of equality of all humans that we hold as moral today. I think it's likely it will benefit some group/ethnicity at the expense of the others. So, a negative resolution for us in the 21st century.

The question can be thought of in terms of incentives: when will the incentive for an established power to fund proxy groups and not lose face in the international community will be seen by said party as more of a disadvantage?

The fact that there are no recent precedents for that from Great Powers is what increases my uncertainty. To me, only Russia or China could disregard international opinion that much, but leaders and priorities will change throughout this century.