TL;DR
An Astral Codex Ten essay argues that calls for AI to rule humans can describe two different positions: AI takeover at human expense, or a handover of some authority to aligned systems in service of human interests. The essay develops the distinction through speculative scenarios about AI, democratic safeguards and political power; it is an argument, not a report of a policy change or an established plan.
Astral Codex Ten has published an essay distinguishing proposals for AI to replace human rule from proposals to delegate authority to aligned AI in the service of people. The distinction matters in debates about AI safety and political power: the author says the two positions are being treated as equivalent, even though one rejects human interests and the other claims to prioritize them.
The essay calls the first position “successionism”: AIs take over, killing people or overriding their interests. The author describes this as “very bad” and says opposing it is central to AI safety. By contrast, “handoverism” means that humans remain the priority while accepting that aligned AI systems might be placed in positions of control if that could serve human goals. The writer says some people working in AI safety, including the author, find this second position plausible.
To explain the case, the author argues that political systems depend on limits to what leaders can do. In the essay’s account, democratic governments are protected in part because workers and soldiers can withhold support from a leader who tries to rule as a tyrant. If increasingly capable AI replaces those groups, the author warns, that constraint could weaken. A small group of officials might then gain power that is difficult to reverse, even where a country formally retains democratic institutions.
The essay considers an alternative in which military AI systems are aligned with democratic principles rather than an individual leader. But that would itself place a foundational role in the hands of AI, the author says. The supplied report then starts a further proposal: allowing an AI to run for president. It introduces a hypothetical 2028 election but ends before completing the scenario, so it does not establish what the remaining proposals are or whether the author endorses a specific implementation.
The Stakes for Democratic Control
The argument is relevant to AI governance and democratic accountability because it asks not only who controls AI, but what happens if human institutions become dependent on systems that can exercise political or military power. If the essay’s scenario came to pass, authority could shift away from voters and human officials, raising questions about who sets a system’s goals and how the public could challenge its decisions.
At the same time, the distinction between delegation and replacement is a conceptual argument, not evidence that AI rule is imminent or that aligned systems can reliably govern. The author’s proposed safeguard depends on AI systems having values that support democracy and human interests. The source does not show how those values would be defined, tested or maintained, nor how citizens could hold such systems accountable.
For readers following AI safety debates, the essay highlights a dispute over what “pro-human” governance could mean. It challenges the assumption that any transfer of authority to AI necessarily rejects human welfare, while making the risks of concentrated power central to the discussion.
As an affiliate, we earn on qualifying purchases.
From AI Takeover to Delegation
The essay responds to criticism of what it calls “AI successionism”. Its author says some critiques have grouped advocates of handoverism with people who want AI to displace humans, while acknowledging that some of the cited essays do make the distinction. The dispute is about how to describe different positions, rather than a newly announced government programme or a change in AI policy.
To make handoverism seem less counterintuitive, the writer compares political institutions to public-service providers. The essay argues that if automated systems can provide coordination or public order better than human officials, delegating some functions could be consistent with prioritizing individuals’ interests. It then develops possible forms of delegation from weaker to stronger, explicitly describing the topic as speculative and science-fiction-like.
The source material provides only part of that sequence. It discusses AI soldiers aligned to democracy and begins a scenario involving an AI presidential candidate, but cuts off mid-example. Those limits mean the full argument and its later proposals cannot be reconstructed from the supplied report.
“True successionism says that AIs should take over, killing humans or steamrolling over their interests.”
— Astral Codex Ten essay author
democratic AI decision-making tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Limits of the Available Proposal
No concrete policy or implementation plan is established in the supplied source. The essay does not specify which AI systems would qualify as aligned, how their commitments would be verified, or what legal limits and oversight would govern them. It also does not explain how people could remove or overrule an AI authority that acted against the public interest.
The text provided is incomplete: it stops during a hypothetical presidential election example. The full set of proposals, the author’s conclusions about them and any responses to the argument are not included. The essay’s claims about how AI might affect labor, military power and democratic checks are predictions, not reported outcomes.
As an affiliate, we earn on qualifying purchases.
The Argument’s Unfinished Scenarios
The immediate next step for readers is to consult the full Astral Codex Ten essay, which may contain the remaining proposals and the author’s conclusions. The supplied excerpt does not give a publication date, announce a response or identify a policy process that will follow.
Further debate is likely to turn on the questions the essay raises: whether AI can be aligned with democratic principles, how authority could be made accountable, and whether delegating power would protect people or create new routes to concentrated control. Those remain questions for discussion; the source reports no agreed answers or scheduled milestones.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is the difference between successionism and handoverism?
Successionism, as defined in the essay, means AI takes power while killing people or overriding human interests. Handoverism means humans remain the priority but may delegate some control to aligned AI if that is judged to serve human goals.
Does the essay call for AI to take over government?
The essay discusses possible forms of AI authority, including AI systems aligned with democracy and the start of a hypothetical AI presidential candidacy. The supplied excerpt is incomplete and does not establish a specific plan or policy call.
Why does the author connect AI with democratic safeguards?
The author argues that workers and soldiers can constrain leaders by refusing to support authoritarian rule. The essay warns that if AI replaced those groups, that safeguard could weaken. This is a speculative argument, not a confirmed account of current events.
What remains unknown about the proposals?
The source does not explain how alignment would be tested, how AI officials would be held accountable, or how the public could reverse a delegation of authority. The excerpt also cuts off before the author completes the list of proposals.
Source: rss
