---
aliases:
  en:
  - AI control research
  ja:
  - AI制御
concept_id: ai_control
preferred:
  en: AI control
  ja: AIコントロール
relations:
  relations:
    broader:
    - ai_safety
    prerequisite:
    - misalignment
status: active
---

# AI control

An AI-safety approach that aims to prevent unacceptable outcomes from powerful AI systems even when those systems are misaligned and intentionally try to subvert the safety measures applied to them.

## broader

- [AI safety](../terminology/ai_safety.txt)

## prerequisite

- [misalignment](../terminology/misalignment.txt)

## Documents: about

- [Control が生存確率を押し上げる世界線は狭い](../documents/akira.control-has-narrow-survival-credit.txt)
- [AIコントロールは時間を稼げても、超知能に対する長期的な安全計画としては頼れない](../documents/greenblatt.control-can-buy-time-but-is-not-a-superintelligence-safety-plan.txt)

## Documents: mentions

- [Untrusted monitoring は共謀や欺瞞によって破られうる](../documents/akira.untrusted-monitoring-can-fail-by-collusion-or-deception.txt)
- [クランチタイムではAIの労働力を防御用途に振り向け、移行を大幅に遅らせる](../documents/cotra.crunch-time-redirect-ai-labor-and-slow-transition.txt)
- [control safety caseには、能力の十分な引き出しと、評価・外挿の保守性が必要になる](../documents/korbak_et_al.control-safety-cases-depend-on-elicitation-transfer-and-extrapolation.txt)

[Corpus index](../index.txt)