LessWrong analysis finds 3 factors that built democracies — and AI threatens them

In a detailed LessWrong post, researchers Sophia Gore, Towards_Keeperhood, and Arslan tackle a question rarely asked by technologists: how did we get to democracy? Until the mid-19th century, almost no one lived under democratic rule. Today, roughly half the world's countries and billions of people are democracies — despite the simultaneous rise of state surveillance, propaganda, and military control. The team categorizes democratic transitions into three clusters: settler-led population reconstitution (e.g., US, Australia), externally driven regime change after military defeat, and gradual domestic shifts. Their literature review and case studies highlight three factors that raise and sustain democratic levels: strong civil society, robust rule of law, and institutionalized political parties.

Critically, the authors note that the 2010s may have been the global peak of democracy, with backsliding already visible. They argue that this equilibrium is fragile and could be disrupted by advanced AI, which may amplify surveillance, propaganda, and power concentration. The post is aimed at the EA and AI safety communities, urging them to understand these dynamics to anticipate how AI might alter political systems. While the team admits they are not experts, their synthesis provides a sobering lens for tech professionals used to thinking about AI's impact on jobs and safety, but not on the very structure of governance.