AI researcher argues democratizing advanced AI crucial to protect democracy

The author opens with a restatement of their core position: frontier AI development should be paused, and any alternative strategies discussed should be understood as contingency planning rather than first-choice paths. The piece then identifies what the author calls post-alignment risks—catastrophic risks that remain even after the alignment problem is solved. The unifying driver behind many of these risks, according to the author, is that by strong default, ASI would end liberal democracy. The author defines liberalism as a system in which people have individual rights, autonomy, and the ability to choose their own destiny, and argues that liberalism is an important force protecting human welfare because when people are free, they are reasonably good at making their lives better of their own volition. The author then catalogs several post-alignment risks: AI-empowered terrorism, coups, permanent dictatorships, and concentration of power. The author acknowledges that these risks already exist today and existed 20 years ago, but notes they are mitigated by the fact that power is relatively evenly distributed across people—the most powerful person in the world does not have an extraordinary advantage over the tenth-most-powerful person. The author's implication is that ASI, by concentrating vast capability in fewer hands, would eliminate this mitigating factor. The article proposes democratizing ASI as a contingency strategy but does not elaborate on what such democratization would entail or how it would be achieved.