[taftster]

Well, it should be the approver of the PR, not the author (AI slop or human slop) that is accountable. I don't ever want an AI to auto-approve a PR (or maybe only for very small things, like dependency-bot kind of tasks).

Not saying that's how it's done, in terms of accountability. The skin-in-the-game thing is hopefully still present, even with AI. But you're right, there's risk.