# Contested concept: control

_**What you are reading**: one contested-concept page in Deliberus's deliberation graph. Deliberus maps human reasoning — including claims that are contested, mistaken, or extreme — so that disagreement can be examined precisely. A claim's presence here is not an endorsement: the graph records that someone argued it, how it connects to other reasoning, and how far scrutiny has gotten. The metadata on this page describes the state of that scrutiny, not the truth of the claim._

---

- **Contestedness**: semantic
- **Diagnosis**: The word names two different objects in this debate, and an argument for one is easily heard as an argument for the other.
- **Canonical**: `/concept/control` (data: `/concept/control.json`)

## Senses in use
- **control over what a machine is permitted to do** — used in 3 claim(s)
- **control over which humans can reach a capability** — used in 3 claim(s)

## Claims using this term
- [Anthropic's framework prioritises catastrophic misuse and loss of control, and escalates safeguards as assessed capability rises: weight se…](/claim/claim_aig_anth_framework.md) _(sense: control over which humans can reach a capability)_
- [Machine authority should be bounded: least privilege, sandboxed network and tool access, and human approval for high-consequence actions.](/claim/claim_aig_bound_machines.md) _(sense: control over what a machine is permitted to do)_
- [Bounding an agent's authority stops an AI from exceeding what its operator allowed; it does not stop a malicious operator from deliberately…](/claim/claim_aig_bounded_scope.md) _(sense: control over what a machine is permitted to do)_
- [Centralized control can reduce misuse through access control, monitoring and revocation, and it lowers the latency of coordinating a respon…](/claim/claim_aig_central_benefits.md) _(sense: control over which humans can reach a capability)_
- [Once a model crosses a dangerous-capability threshold, stronger custody (secured weights, mediated access, monitoring, revocation) is the r…](/claim/claim_aig_custody_warranted.md) _(sense: control over which humans can reach a capability)_
- [AI governance should aim to preserve broad human control through bounded machine authority, distributed human capability, plural institutio…](/claim/claim_aig_goal.md) _(sense: control over what a machine is permitted to do)_