{"slug":"accuracy-robustness-security-testing","id":"AIPT-CTL-0001","title":"Accuracy, robustness, fairness and security testing","kind":"technical","kind_label":"Technical measure","purpose":"Verifies before release, and again after material change, that a system performs to declared accuracy levels, behaves consistently across groups and conditions, and resists the attacks specific to AI.\n","description":"A test plan defines the metrics, datasets, thresholds and acceptance criteria for each trustworthiness property the risk assessment flagged: accuracy and calibration, performance across demographic and operational subgroups, resilience to noisy or out-of-distribution input, and security against data and model poisoning, evasion and extraction. Tests are run by people independent of the developers where practical, results are compared against the criteria and failures block release until fixed or formally accepted. Declared performance figures in user documentation are drawn directly from the latest report, and the plan is re-run on retraining or a change in operating context.\n","owner_role":"Quality or testing lead","frequency":"on_material_change","frequency_label":"At launch and on material change","risk_subdomains":["7.3","1.3","2.2","1.1"],"evidence":[{"type":"evaluation_report","title":"Pre-release test report","description":"Metrics, subgroup results, robustness and security findings against acceptance criteria."},{"type":"procedure","title":"Test plan and acceptance criteria","description":null},{"type":"approval_record","title":"Release test sign-off","description":null}],"framework_references":[{"framework":"iso_42001","framework_name":"ISO/IEC 42001:2023","reference":"Annex A.6.2.4","note":"Verification and validation.","confidence_level":"high"},{"framework":"nist_ai_rmf","framework_name":"NIST AI RMF 1.0","reference":"MEASURE 2.5, 2.6, 2.7, 2.11","note":null,"confidence_level":"high"},{"framework":"iso_27001","framework_name":"ISO/IEC 27001:2022","reference":"Clause 8.1; Annex A 8.29 Security testing in development","note":null,"confidence_level":"medium"},{"framework":"mitre_atlas","framework_name":"MITRE ATLAS","reference":"AML.M0008 Validate ML Model, AML.M0003 Model Hardening, AML.M0015 Adversarial Input Detection","note":null,"confidence_level":"medium"},{"framework":"owasp_llm_top10","framework_name":"OWASP Top 10 for LLM Applications","reference":"LLM04 Data and Model Poisoning","note":null,"confidence_level":"medium"}],"obligations":[{"slug":"eu-ai-act-risk-management-system","title":"Establish a risk management system for high-risk AI","relationship":"supports","note":"Testing confirms that mitigations work.","confidence_level":"high","is_binding":true,"policy":"eu-ai-act","jurisdiction":"eu","url":"https://aipolicytracker.org/obligations/eu-ai-act-risk-management-system"},{"slug":"australia-vaiss-testing-human-control-transparency","title":"Test and monitor systems, enable human control, and be transparent with users (guardrails 4 to 6)","relationship":"supports","note":"Guardrail 3: pre-deployment testing.","confidence_level":"high","is_binding":false,"policy":"australia-voluntary-ai-safety-standard","jurisdiction":"australia","url":"https://aipolicytracker.org/obligations/australia-vaiss-testing-human-control-transparency"},{"slug":"eu-ai-act-art-50-2-synthetic-content-marking","title":"Providers of generative AI must mark synthetic output as artificially generated in a machine-readable way","relationship":"supports","note":"Tests that the mark is robust and reliable.","confidence_level":"medium","is_binding":true,"policy":"eu-ai-act","jurisdiction":"eu","url":"https://aipolicytracker.org/obligations/eu-ai-act-art-50-2-synthetic-content-marking"},{"slug":"uk-principles-safety-security-robustness","title":"Ensure AI systems are safe, secure and robust throughout their lifecycle","relationship":"satisfies","note":"Security and robustness evidence to show a regulator.","confidence_level":"medium","is_binding":false,"policy":"uk-ai-regulation-white-paper","jurisdiction":"uk","url":"https://aipolicytracker.org/obligations/uk-principles-safety-security-robustness"},{"slug":"uk-principles-fairness","title":"Use AI in ways that are fair and do not discriminate unlawfully","relationship":"satisfies","note":"Disparate-outcome testing across protected characteristics before and after deployment.","confidence_level":"medium","is_binding":false,"policy":"uk-ai-regulation-white-paper","jurisdiction":"uk","url":"https://aipolicytracker.org/obligations/uk-principles-fairness"},{"slug":"india-dpdp-security-and-breach-notification","title":"Implement reasonable security safeguards and notify breaches","relationship":"supports","note":"Reasonable security safeguards evidenced by testing.","confidence_level":"medium","is_binding":true,"policy":"india-dpdp-act","jurisdiction":"india","url":"https://aipolicytracker.org/obligations/india-dpdp-security-and-breach-notification"},{"slug":"eu-ai-act-accuracy-robustness-cybersecurity","title":"Achieve appropriate accuracy, robustness and cybersecurity","relationship":"satisfies","note":"Declared accuracy metrics, robustness and AI-specific security testing come from the release test programme.","confidence_level":"high","is_binding":true,"policy":"eu-ai-act","jurisdiction":"eu","url":"https://aipolicytracker.org/obligations/eu-ai-act-accuracy-robustness-cybersecurity"},{"slug":"us-new-york-city-local-law-144-bias-audit","title":"Employers and employment agencies must obtain an independent bias audit before using an automated employment decision tool","relationship":"satisfies","note":"Disparate-impact testing performed by an independent auditor on the statutory metrics.","confidence_level":"high","is_binding":true,"policy":"us-new-york-city-local-law-144-automated-employment-decision-tools","jurisdiction":"us-new-york","url":"https://aipolicytracker.org/obligations/us-new-york-city-local-law-144-bias-audit"},{"slug":"eu-ai-act-art-55-systemic-risk-cybersecurity","title":"Providers of systemic-risk GPAI models must secure the model and its infrastructure","relationship":"satisfies","note":"Security testing of the model and its serving infrastructure.","confidence_level":"medium","is_binding":true,"policy":"eu-ai-act","jurisdiction":"eu","url":"https://aipolicytracker.org/obligations/eu-ai-act-art-55-systemic-risk-cybersecurity"},{"slug":"singapore-mgf-operations-management","title":"Manage data quality, model development and monitoring across the lifecycle","relationship":"supports","note":"Repeatability and robustness evidence before release.","confidence_level":"medium","is_binding":false,"policy":"singapore-model-ai-governance-framework","jurisdiction":"singapore","url":"https://aipolicytracker.org/obligations/singapore-mgf-operations-management"},{"slug":"us-colorado-ai-act-developer-reasonable-care","title":"Developers must use reasonable care to avoid algorithmic discrimination","relationship":"supports","note":"Disparate-impact testing across protected classes.","confidence_level":"high","is_binding":true,"policy":"us-colorado-ai-act","jurisdiction":"us-colorado","url":"https://aipolicytracker.org/obligations/us-colorado-ai-act-developer-reasonable-care"},{"slug":"us-new-york-city-local-law-144-publish-audit-summary","title":"Employers and employment agencies must publish a summary of the bias audit results","relationship":"supports","note":"Produces the results being summarised.","confidence_level":"high","is_binding":true,"policy":"us-new-york-city-local-law-144-automated-employment-decision-tools","jurisdiction":"us-new-york","url":"https://aipolicytracker.org/obligations/us-new-york-city-local-law-144-publish-audit-summary"},{"slug":"us-nist-ai-rmf-measure","title":"Measure and test trustworthiness characteristics (Measure)","relationship":"satisfies","note":"Metrics and test methods for each trustworthiness characteristic.","confidence_level":"medium","is_binding":false,"policy":"us-nist-ai-rmf","jurisdiction":"us","url":"https://aipolicytracker.org/obligations/us-nist-ai-rmf-measure"},{"slug":"us-texas-responsible-ai-governance-act-traiga-unlawful-discrimination-prohibition","title":"Developers and deployers must not use AI with the intent to unlawfully discriminate against a protected class","relationship":"supports","note":"Fairness testing evidence supports the absence of intent.","confidence_level":"medium","is_binding":true,"policy":"us-texas-responsible-ai-governance-act-traiga","jurisdiction":"us-texas","url":"https://aipolicytracker.org/obligations/us-texas-responsible-ai-governance-act-traiga-unlawful-discrimination-prohibition"},{"slug":"us-omb-m-25-21-high-impact-ai-practices","title":"Apply minimum risk-management practices to high-impact AI","relationship":"supports","note":"Pre-deployment test evidence agencies can reuse.","confidence_level":"medium","is_binding":true,"policy":"us-omb-m-25-21","jurisdiction":"us","url":"https://aipolicytracker.org/obligations/us-omb-m-25-21-high-impact-ai-practices"}],"related_controls":["ai-risk-assessment","genai-red-team-testing","post-market-monitoring-and-drift-detection","model-release-and-change-management-gate"],"review_status":"verified","confidence_level":"high","last_verified_at":"2026-09-26","reviewed_by":"Bhaskar Bhatt","updated_at":"2026-10-15T06:46:22+00:00","url":"https://aipolicytracker.org/controls/accuracy-robustness-security-testing","context_url":"https://aipolicytracker.org/controls/accuracy-robustness-security-testing.md"}