My current favorite is #4, but that's a transitionary state.  Corporations are neither benevolent nor malevolent, they are self-sustaining agent-like actors with somewhat opaque causality behind their actions.  Much like AI, for that matter.  They will appear benevolent as long as there are sufficient opposing forces that they need the illusion, and then reveal their disinterest when it's no longer true.

I'd bet on #4.

Or an alternative combo of 3 and 4: e.g., the AI-empowered corporations continue to gain astronomical wealth until they're largely more powerful than national governments. The process of mergers, acquisitions, and takeovers from AI companies leads to megacorporation oligopolies that control geographic and cyber empires. Incrementally, the economic power of large AI giants begins to encompass geopolitical power--akin to the historical shift from empires to city-states to nation states--and gradually company security debarments become armies, and company leadership, functional authoritarian ASI governments.