[Mindey]

The idea of founding "OpenAI" was: Make AI an extension of human wills in the spirit of liberty, as broadly and evenly distributed as is possible safely. The idea was that if we empower people evenly, any evil people powered by AI would not be able to do a 51% attack on the people empowered by AIs collectively.

Now, suppose you "Open" the OpenAI, by giving the information about the model and data, all of that -- the only ones who would empower by that kind of openness -- is the very large corporations and the governments of powerful countries rather than people evenly.

So, while distributing the AI power evenly in the society ("democratize AI") was the original motivation of OpenAI group, and founding it as a non-profit may have motivated researchers and capable engineers join the effort, the reason why it isn't staying "open" is simply, because they had not yet figured out how to "Empower people evenly", rather than empower large governments and corporations through such openness.

[Mindey]

Additionally, Microsoft has for a long time been embraced by the world's governments as a solution to dealing with the official matters, most of the world's governments are users of it with access to the source code (Government Security Program (GSP)), and trust Microsoft to a high degree. Such trust may be important as a kind of justification for the world's governments that this AI is safe hands, however, obviously, every larger government is strategizing in their back-offices the implications of such AI to their security... and when AI becomes a strategic weapon, you bet it can't easily be open.