Transparency & Documentation
High-Risk AI Deployer Instructions
What this control does
Provide clear, accessible instructions covering intended use, limitations, performance, oversight, maintenance and logging.
How to implement
For high-risk AI providers, assign an owner to the instructions supplied with each system version. Explain intended purpose, capabilities, limitations, relevant performance information, input-data expectations, oversight, maintenance and logging in terms the deployer can use. Test whether a representative user can interpret an output and recognize a limitation. Make the instructions accessible and distribute updates with the affected release. Accessible instructions support part of an accessibility program; they do not by themselves establish that the whole system meets every applicable accessibility requirement.
Suggested timing and triggers
Before release; with relevant version or capability changes; periodic usability and accessibility review set by the organization.
Evidence examples
Approved, versioned instructions and distribution records Traceability to tested capabilities and known limitations User walkthrough and accessibility review findings Update notices and resolved documentation issues
How to check this control
Use a representative task to test whether a deployer can interpret the output, locate a limitation and follow the oversight instructions. Compare the instructions with the released system and investigate any mismatch.
Related EU AI Act obligations
Article 13
Provide Transparency and Instructions for Use to High-Risk AI Deployers
Article 13(1)
Design High-Risk AI for Sufficient Operational Transparency
Sub-obligation of Article 13: Provide Transparency and Instructions for Use to High-Risk AI Deployers
Article 13(2)-(3)
Provide Clear and Complete High-Risk AI Instructions for Use
Sub-obligation of Article 13: Provide Transparency and Instructions for Use to High-Risk AI Deployers
Article 16(l)
Ensure High-Risk AI Meets Applicable Accessibility Requirements