[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"raw-en-risks":3},"---\ntitle: AI Risk Management\ndescription: IQ:NS models seven risk dimensions — fairness, safety, security, privacy, transparency, accountability, and reliability.\nlang: en\nnavigation:\n  section: ontologies\n  label: Risk\n  order: 30\n---\n\n# AI Risk Management\n\n## Why risk structure matters\n\nAI systems fail in specific, predictable ways. Different frameworks categorise those failures differently. IQ:NS normalises them into seven dimensions drawn from the major standards, so risk concepts map consistently across every framework you care about.\n\n---\n\n## The seven dimensions\n\n### 1. Fairness & Bias\nModels discriminate against protected groups in high-impact decisions. Covered by EU AI Act, ECOA, Fair Lending Act, NIST Measure 2.6, ISO 24027.\n\n### 2. Safety\nAI causes or contributes to physical harm, financial loss, or misdiagnosis. Covered by EU AI Act Annex III, NIST, ISO 42001, sector standards.\n\n### 3. Security\nAdversarial attacks, model theft, data poisoning, prompt injection. Covered by OWASP LLM Top 10, MITRE ATLAS, Google SAIF, NIST AI 100-1.\n\n### 4. Privacy\nProcessing personal data without valid legal basis, violating minimisation principles, automated decisions without safeguards. Covered by GDPR Articles 22\u002F35\u002F36, EU AI Act.\n\n### 5. Transparency & Explainability\nUsers don't know they're interacting with AI. Decisions can't be explained. Covered by EU AI Act Chapter IV, NIST Govern 4.1, ISO A.8.3.\n\n### 6. Accountability & Oversight\nNo one owns AI decisions. Approval processes don't exist or get bypassed. Covered by all major frameworks.\n\n### 7. Reliability & Robustness\nModels drift, degrade, or behave unpredictably. Covered by ISO 24029, NIST, monitoring requirements across frameworks.\n\n---\n\n## How IQ:NS manages risk\n\nEach risk dimension maps to specific obligations across every applicable framework. The ontologies capture:\n\n- Which frameworks define this risk category\n- Where the definitions align and where they diverge\n- What controls each framework requires\n- How concepts relate across standards\n\nThis means a single query can show you every obligation related to, say, fairness — across EU AI Act, NIST, ISO, and sector-specific rules — in one structured view.\n\n---\n\n[View the standards coverage](\u002Fstandards) · [Explore the ontologies](https:\u002F\u002Fgithub.com\u002Fiqns-org\u002Fontologies)\n",1776235631411]