https://taxonomy.eticas.ai/risk/reliability
Maturity: established
The risk that an AI system produces false, fabricated, or misleading outputs (hallucinations), spreads inaccurate or deceptive information (misinformation), or delivers inconsistent results across similar inputs and contexts. Such failures undermine trust, reduce system dependability, and can lead to harmful or misguided decisions.
Also known as: Reliability & Manipulation · Validity and Reliability
Applies to: ALL
Lifecycle stages: Pre Processing, In Processing, Post Processing
| Framework | Concept |
|---|---|
| EU AI Act (Regulation 2024/1689) | Article 15 — accuracy, robustness and cybersecurity |
| ISO/IEC 42001:2023 — AI Management System | AI system verification and validation |
| AIUC-1 — AI Underwriting Company Standard | Reliability domain |
| Framework | Concept |
|---|---|
| NIST AI 600-1 — Generative AI Risk Profile | Confabulation |
| NIST AI 600-1 — Generative AI Risk Profile | Information Integrity |
| NIST AI Risk Management Framework (AI 100-1) | Valid & Reliable |
| OECD AI Principles | Robustness, security & safety |
| Framework | Concept |
|---|---|
| MIT AI Risk Repository | Lack of capability or robustness |
| MIT AI Risk Repository | Misinformation |
| AIR 2024 / AIR-Bench 2024 | System & Operational Risks → Operational Misuses |
| IBM AI Risk Atlas | Output → Robustness + Inference → Accuracy |