[shekhirin]

I think there's two points to unpack:

1. As you said, it was compute-rich before, and native token-rich now. I see it that way: the difference is that your richness is now a part of the protocol which allows it to e.g. punish bad actors (aka slashing) which isn't possible in PoW scenario where attacks can continue indefinitely.

2. Validators don't solely decide rules of execution: full nodes do. If you, as a validator, try to break the rules (double spend, incorrect execution, etc.), full nodes won't accept this new world view and discard it. So providers like Infura which most users (currently) use will continue working as expected and the canonical chain won't be damaged.

[__alexs]

Who defines what a "bad" actor is?

[DennisP]

Objective protocol rules. If you send staking transactions that conflict with each other in certain specific ways, then you get slashed.