AI is possibly the “last invention” mankind would need to make, and the rewards are incalculable, which is also true of the possible negative consequences. This does not mean there should be a moratorium on the technology.

Additionally, the genie is already out of the bottle, and the US would not be able to legitimately enforce some type of technology embargo inside foreign nations such as China and Russia, nor would the US be able to prevent its development, which would serve only to put us at a disadvantage.

There is also the issue that such an invention basically would eliminate the needs for most service jobs in primarily the financial industry, but other industries as well, such as information technology and research & development.

In order to prevent such massive power from being concentrated in too few hands, it needs to be handled much more like the internet is, with no singular country or corporation (I know you are thinking of google, but I am thinking IBM) in control of all the processing power. This ideal of democratization is the basis for the OpenAI project.

This means, much like how the DNS system for the internet is being handed over to an international consortium, so too should the management of a distributed AI system. It could feasibly become a part of the UN mandate, which would also allow it to be used to solve some of the most intractable problems brought forth to the UN by member countries while also drastically increasing the desirability of being a UN member state and the importance of international cooperation as we tackle global problems such as climate change and endless war.

So the answer is Democracy where no individual organization is able to exert unilateral control. The most obvious would be for the G8 countries to create an division of AI affairs, or Department for AI integration, or Section8, 

UNAIC (UN AI Counsel)

or AI Consortium, where multiple AIs would be working with human representatives, to head off the AI war., I know that sounds like the beginning of a porno, but it’s really important for hashing out the ethics and policies regarding deployment of AI technologies on a commercial scale and to prevent fascists from being able to use “AI” as a cover for their non-AI driven operations.

For example, imagine trump being the president and having control over an AI, or the ability to claim any decision was the result of an AI. A dark vision indeed and underlines the importance of this technology being distributed in the same way the internet is.

Oh also, the primary reason for it to be an international governmental consortium (not just consortium of US corporations) designed such that it has multiple AIs as well as human representative counterparts, is in order to prevent AI war, which is where an individual organization or country could be tempted to attach another via indirect means. AI war would not be something that can be de-escalated due to the rapid response nature (think Skynet) thus every effort to prevent the possibility of MAD AI (mutually assured destruction by one or multiple defence artificial intelligences)