Normally, online games are like social networks (silos, walled gardens where the rule makers can change them for their own profit, slowly destroying the Game and making it a Zero-sum-game, just like other social networks or the great Fiat ponzi. When the rules of a game constantly change, your different investments can collapse catastrophically because it's like the measuring instruments for the different projects you build suddenly become imprecise and failures creep through your plans)

I thought it only was like that because the cost of developing and running an online game through servers is normally enormous. But I refuse to be bored all my life. I thought a cheaper method could disincentivize the abusive behavior of making pay-to-play oriented rule-changes to bleed your consumers (the players) without any consent

In the case I presented you, the advantages of using AI is that it could understand relations and interactions that could take a malicious human an excruciatingly big amounts of time to taint. Changes that are despotically conceived by a human could just provoke an irreversible chaos to the game.

But in this case because AI can potentially become mostly Open source, anyone could just fork the game with other AI models of similar nature and keep the previous "stable economy" thriving.

A good example of the collapse of an stable Game economy is in the article of Gigi about Diablo 2 and the stone of Jordan: