There is this excellent prompt from @ahall_research in his article on how do we design agents that genuinely serve democratic self-governance and how to keep them accountable. It is a completely different lens than what folks over in AI safety thinking. Article: