The easiest pathway to control is through executive power

According to LessWrong contributor djbinder, artificial intelligence researchers often focus on elaborate loss-of-control scenarios—rogue AIs developing nanotechnology, coordinated bioweapon attacks, or mass persuasion campaigns. But djbinder argues the more plausible path to AI-enabled power concentration runs through something simpler: existing government structures. The post contends that in both the United States and China, concentrated executive power already exists—particularly over military and security apparatus. During a rapid artificial intelligence transition, national security concerns would naturally expand executive authority. This could enable gradual, quasi-legal consolidation of permanent control without requiring obviously illegal moves that subordinates would refuse. Djbinder notes this follows a pattern seen in modern authoritarian transitions—Hitler, Putin, Orbán, and Erdoğan all gradually hollowed out democratic institutions without military coups. The challenge, he argues, is that most democratic institutions operate on slow timescales—elections years away, court cases taking months—too slow to check rapid executive moves. Once power is consolidated, increasingly powerful A-I tools could help make that control permanent. The post concludes that designing institutions resilient to this scenario is harder than most AI safety discussions suggest, but remains crucial.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton