https://taxonomy.eticas.ai/risk/autonomy-agency
Maturity: emerging
The risk that an AI system undermines human agency or operates with insufficient human control. This includes over-reliance on AI outputs, automation bias, deskilling, and — for agentic AI systems — failure modes arising from multi-step autonomous actions, tool use, or coordination with other AI agents.
This category is emerging. Assessment methods are still being developed. Definitions and subcategories may evolve.
Also known as: Autonomy & Human Agency · Human-AI Interaction
System type: ADM and LLM systems
Lifecycle stages: Post Processing
| Framework | Reference |
|---|---|
| EU AI Act (Regulation 2024/1689) | Article 14 — Human oversight |
| AIUC-1 — AI Underwriting Company Standard | Prevent unauthorized AI agent actions + C.7 Human review |
| ISO/IEC 42001:2023 — AI Management System | Human oversight (lifecycle controls) |
| Council of Europe Framework Convention on AI (CETS No. 225) | Articles 7–8 — Human agency and oversight (implied) |
| NIST AI 600-1 — Generative AI Risk Profile | Human-AI Configuration |
| NIST AI Risk Management Framework (AI 100-1) | GOVERN 3 — Human-AI configurations + MANAGE 1 risk-prioritised oversight |
| OECD AI Principles | Principle 1.2 — human agency and oversight |
| Framework | Reference |
|---|---|
| MIT AI Risk Repository | Human-Computer Interaction |
| W3C Data Privacy Vocabulary — AI Extension | Human Oversight |
| IBM AI Risk Atlas | Non-technical → Value alignment + Output → Anthropomorphisation |