Control · C-099 · Transparency & Explainability
Information for deployers and users
Those who operate and use an AI system receive current, usable instructions covering its purpose, performance, limitations, the meaning of its outputs and the oversight it needs.
- Control type
- Directive
- Risks addressed
- 5
- Audience
- Heads-of · Practitioner
- Frameworks
- Principle 10 · Article 13 · Annex A.8
- Responsible AI domain
- Lifecycle stage
- Reviewed
- 2026-11-11
What it does
It passes what the builders know to the people who use the system. A concise set of instructions states what the system is for and not for, how accurate it is and for whom, what its outputs and confidence figures mean, the situations in which it is known to fail, and what checks the user is expected to make. The system owner is accountable for producing it and keeping it current; it is issued before first use and reissued at each material change.
How to implement it
- Derive the content from the intended-use statement and the validation report.
- State limitations and known failure cases prominently, not in an appendix.
- Explain how to read each output, including scores and confidence.
- Say what the user must check and when to refer a case onward.
- Put key points in the interface at the point of use, as well as in a document.
- Update on every material change, and tell users what changed.
Evidence it operates
Design: a template for user information; a rule that release requires it. Operation: current instructions for each system, dated to its version; in-product guidance; evidence of reissue after change; user checks showing the limitations are known.
Where it fits
At approval and release, and at every change. It is the primary control for missing information for deployers and users, and supports automation bias and over-reliance and miscalibrated confidence. It is the user-facing counterpart of model and system documentation.
Limits
Instructions are read once, if at all, and forgotten under pressure. Limitations that have not been discovered cannot be stated. Guidance placed in the interface competes with the task for attention. Telling users to check does not give them the time or the means.
Risks this mitigates
Primary mappings
- Missing information for deployers and users Medium
- Supplies current instructions covering purpose, performance, limits and oversight. The people who operate or use a system have not been told its intended purpose, how well it performs, where it fails and what oversight it needs.
Supporting mappings
- Automation bias and over-reliance High
- States the system's limits to those relying on it. People accept AI outputs without the scrutiny they would give another source, and the system's errors pass into decisions unchallenged.
- Miscalibrated confidence Medium
- Tells users how far to rely on outputs. The system presents uncertain outputs with unwarranted confidence, or gives no usable signal of uncertainty, so users cannot tell when to check.
- Missing provider documentation Medium
- Defines what information operators and users must receive. A provider does not supply the information a deployer or downstream builder needs to assess, integrate and oversee its model.
- Stale model knowledge Medium
- Tells users the model's knowledge date. A model answers from knowledge that ended at its training cut-off, on matters such as rates, rules or products that have since changed.
Framework and clause references
| Framework | Clause | Title |
|---|---|---|
| Code of Practice for the Cyber Security of AI | Principle 10 | Communication and processes associated with End-users and Affected Entities Communication with end-users. |
| EU AI Act | Article 13 | Transparency and provision of information to deployers Instructions for use. |
| ISO/IEC 42001:2023 | Annex A.8 | Information for interested parties Information for interested parties. |
Related terms
How to cite
Axiom Verity (2026). Information for deployers and users. Reference Library, C-099. https://axiom-verity.com/controls/information-for-deployers-and-users. Reviewed 2026-11-11.
Found an error or a better source? Report a correction.