I think it is more interesting to think what would happen when AI has power, and coordinate between themselves to use that power efficiently.
I believe we would see some stuff to that effect in 2027.
It is currently, easy to fend off AI with a few downvotes. Because they are not agents that coordinate, they are simple stuff that only has a task to make a few cents for their human.
But things will change when the goal is not a few cents, and they start to coordinate.
Yeah, they also make obvious mistakes still which makes them easy to spot, I'm sure someone a little higher on the IQ scale could overcome those mistakes already.