CLUSTER / THE DECISION MAP
High-risk requirements: risk management, data, documentation, oversight, accuracy and robustness.
The Section 2 requirements for high-risk systems, one guide per requirement, and how evidence chains are built.
IN THIS CLUSTER
13 guides.
What requirements must high-risk AI systems meet?
Lead overview of the high-risk requirements in Articles 9 to 15: risk management, data, documentation, record-keeping, transparency, human oversight and accuracy, robustness and cybersecurity.
Read guide ↗CLUSTER ARTICLEWhat risk management system does Article 9 require?
Article 9 explained: a continuous, iterative, documented risk management process across the lifecycle of a high-risk AI system.
Read guide ↗CLUSTER ARTICLEWhat data and data governance does Article 10 require?
Article 10 explained: data governance practices, dataset quality criteria, bias examination and the Article 4a treatment of special categories of personal data.
Read guide ↗CLUSTER ARTICLEWhat technical documentation does Article 11 require?
Article 11 and Annex IV explained: the contents, currency, retention and simplified documentation treatment for high-risk AI systems.
Read guide ↗CLUSTER ARTICLEWhat record-keeping and logging does Article 12 require?
Article 12 explained: automatic event recording, log contents, retention and the link between logs, post-market monitoring and incident handling.
Read guide ↗CLUSTER ARTICLEWhat transparency and information must high-risk systems give to deployers?
Article 13 explained: instructions for use, capabilities and limitations, oversight and monitoring information, and the deployer's ability to interpret output.
Read guide ↗CLUSTER ARTICLEWhat human oversight does Article 14 require?
Article 14 explained: oversight design, the ability to understand, monitor, intervene and halt, and the special position of biometric identification systems.
Read guide ↗CLUSTER ARTICLEWhat accuracy, robustness and cybersecurity does Article 15 require?
Article 15 explained: appropriate accuracy, robustness against errors and adversarial threats, and cybersecurity protection across the lifecycle.
Read guide ↗CLUSTER ARTICLEHow should testing and validation be planned for high-risk systems?
Testing and validation across Articles 9 and 15: predefined metrics and thresholds, representative test data, adversarial testing and lifecycle retesting.
Read guide ↗CLUSTER ARTICLEHow should bias be detected and corrected in high-risk systems?
Bias detection and correction under Articles 9, 10 and 4a: examination methods, performance differences, correction measures and the limits of special category data processing.
Read guide ↗CLUSTER ARTICLEWhat is the difference between high-risk requirements and actor obligations?
Requirements in Articles 9 to 15 distinguished from obligations in Articles 16 to 27: who must do what, and how the two layers are evidenced together.
Read guide ↗CLUSTER ARTICLEHow should evidence chains be built for high-risk requirements?
Evidence chains explained: linking classification, requirements, testing, documentation, conformity and post-market records into one traceable file.
Read guide ↗CLUSTER ARTICLEHow do the high-risk requirements work together in a real system?
A worked, general walkthrough of Articles 9 to 15 applied together, from intended purpose to post-market monitoring, without deciding any specific system.
Read guide ↗