Well, at the moment, AIs don't get to vote. So step 1 is, don't vote them into office (or even into the franchise).

That is, AIs don't get to govern things without being put into governmental power by humans. Don't let the humans be the voters.

That doesn't stop office holders from using them like a magic 8-ball. So step 2 is, vote out office holders that let AIs write laws (or even replies to constituents).

My magic 8-ball says that major corporations are training the LLMs and pushing LLM features into our eyeballs at every opportunity. How do you propose we convince the average voter that our stance on AI is the correct one? If I were a power-hungry corporation, I'd be feeding my LLMs with the texts of laws as I want them to written, transcripts of cases as I'd wish they'd been tried, arguments for candidates and policies that I like... so I, an individual with less than a million dollars to my name, do not know how to compete with this.

How do I propose to convince them? I don't need to. As corporations push LLM features into their eyeballs at every opportunity, they'll get the picture.