TL;DR
An Astral Codex Ten essay argues that calls for AI to replace human rule can describe two different positions: AI domination at human expense, and using aligned AI to serve human interests. The author says the latter could become attractive as AI weakens human safeguards against authoritarian rule, but the proposals are speculative and raise unresolved questions about accountability and control.
An Astral Codex Ten essay argues that debate over AI taking political power often conflates two positions: AI replacing humans at their expense, and aligned AI exercising authority to serve human interests. The distinction matters as increasingly capable AI could, the author argues, weaken the human labor and military safeguards that currently constrain would-be dictators.
The author calls the first position “true successionism”: the belief that AIs should take over even if that means killing people or overriding their interests. The essay describes this as incompatible with AI safety. It contrasts that with “handoverism,” which prioritizes human interests but allows that those interests might be better served by aligned AIs holding positions of control. The author says some people in AI safety find this second position plausible, and objects to treating its supporters as proponents of an anti-human takeover.
The argument begins with a concern about political power. In the essay’s account, democratic leaders face limits because they depend on workers to sustain the economy and soldiers to carry out military orders. If advanced AI replaces much of that labor and military role, the author warns, a small group of officials could face fewer practical obstacles to consolidating power. The essay says existing safeguards may be difficult to reproduce once human buy-in is no longer central.
One proposed response is to make AI military systems loyal to democratic principles rather than to a particular officeholder or political faction. The author calls that arrangement a form of handover in itself: AI systems would become a foundation of government by acting according to values such as democracy. The essay then asks whether it might seem preferable for such systems to govern directly if human politicians remain corrupt or manipulate elections. It introduces an AI presidential candidacy as a possible minimum legal change, but the supplied excerpt ends while beginning to describe a hypothetical election.
AI as a Check on Political Power
The argument is relevant to debates about AI safety and democratic accountability because it treats political control as a potential consequence of advanced AI, not only a question of productivity or military capability. If AI systems take over tasks now performed by workers and soldiers, the author argues, institutions could lose some of the practical resistance that helps prevent leaders from acting against democratic rules.
At the same time, the essay’s distinction challenges a simple equation between any AI authority and an anti-human outcome. Its case for handoverism depends on a demanding condition: the systems exercising power must remain aligned with human interests. The source does not establish that this condition can be achieved or maintained. Instead, it frames the issue as a speculative political problem: whether AI authority might sometimes protect people from human rulers, and what safeguards could prevent that authority from becoming a threat of its own.
As an affiliate, we earn on qualifying purchases.
From AI Soldiers to AI Leaders
The essay uses “handover” to describe a spectrum, rather than a single policy. Its starting point is that democratic governments already depend on institutions whose members follow rules and principles rather than simply obeying one individual. The author compares a future in which AI soldiers are aligned with democracy to the present role of human soldiers, while arguing that the AI version could make that safeguard more final.
The next step in the argument is more politically contentious. Even if AI systems protect democratic procedures, the author says, corrupt human lawmakers could still exploit those procedures through propaganda and patronage. That possibility leads to the question of whether an AI should be permitted to seek the presidency. The essay presents these ideas as proposals of increasing strength, and explicitly describes the subject as speculative and science-fiction-like; the available source excerpt does not provide the full set of five proposals.
“True successionism says that AIs should take over, killing humans or steamrolling over their interests.”
— Astral Codex Ten essay author
As an affiliate, we earn on qualifying purchases.
Safeguards and Accountability Remain Open
The source presents a political argument, not evidence that AI systems can reliably govern or remain aligned with people over time. It does not specify how democratic values would be defined, who would verify an AI’s conduct, or how citizens could remove or overrule a system holding political or military power. Those questions are central to whether the proposed arrangements could protect human interests rather than merely assert that they do.
The supplied material also ends partway through a hypothetical presidential election scenario. It therefore does not show the full five-proposal sequence, the author’s complete treatment of an AI candidate, or any responses to the argument. No enacted constitutional change or real-world AI candidacy is reported in the source.
As an affiliate, we earn on qualifying purchases.
The Debate Turns to Political Design
The immediate next step is continued debate over the difference between AI replacing human interests and AI being assigned authority in their service. The source does not announce a policy proposal, vote, or government decision. It offers a framework for considering possible political arrangements as AI capabilities grow.
Further reporting or the complete essay would be needed to assess the remaining proposals and any practical safeguards the author discusses. For readers, the key questions are whether human institutions can preserve democratic checks as AI takes on more economic and military work, and what accountability would apply if an AI were given direct governing authority.
AI presidential election simulation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is handoverism?
Handoverism is the essay’s term for prioritizing human interests while allowing that aligned AI might exercise some political control in order to serve them.
How does the essay define successionism?
The author uses true successionism for AI taking over in ways that kill people or override their interests. The essay distinguishes this from AI authority that is intended to benefit people.
The essay argues that democratic leaders are constrained partly by their dependence on workers and soldiers. If AI replaces those roles, the author warns, leaders could face less resistance to consolidating power.
Does the essay report a plan to elect an AI president?
No. It raises an amendment allowing an AI to run for president as a hypothetical possibility. The supplied excerpt does not report an actual candidacy or legislative proposal.
What remains uncertain about AI governance?
The source does not establish how AI alignment could be tested, how systems would be held accountable, or how people could challenge political decisions made by AI. It presents these questions within a speculative argument.
Source: rss
