Article 13 — Transparency and Information to Deployers
Article 13 requires that high-risk AI systems are designed to be sufficiently transparent for deployers to interpret outputs and use the system appropriately. Providers must supply comprehensive instructions of use.
Art. 13(1)
System sufficiently transparent for deployers to interpret outputs
The deployer guide is a comprehensive transparency document covering intended purpose, capabilities, limitations, and interpretation guidance. It depends on five upstream artifacts ensuring all transparency claims reflect actual system behavior.
Art. 13(3)(a)
Provider identity and contact details
Provider Identity and Contact section is a required template field. This information is also referenced in DEPLOY.md for operational contact purposes.
Art. 13(3)(b)
Intended purpose, accuracy/robustness levels, known limitations
Three dedicated sections cover system capabilities (from ARCH), performance expectations (from QA_REPORT and MODEL_EVALUATION), and known limitations. Cascade invalidation ensures these sections update when upstream metrics change.
Art. 13(3)(d)
Human oversight measures
The human oversight section documents what oversight is required, how deployers should implement it, and what decisions the system should not make autonomously. This directly addresses Article 14 requirements from the deployer perspective.