Transparency Standard
Standard ID: GCAIS-STD-002Table of Contents
1. Scope
This standard establishes disclosure requirements for organizations deploying AI systems in contexts where users or affected persons interact with or are subject to the outputs of those systems. It applies to all AI systems in production deployment where the system's involvement in generating outputs or informing decisions is not self-evident to affected persons.
Requirements apply to the deploying organization and govern what information must be disclosed, to whom, in what form, and at what point in the user interaction. Requirements do not mandate disclosure of proprietary technical information that would constitute a material competitive disadvantage, except where such disclosure is required for meaningful human oversight.
This standard does not supersede applicable legal disclosure requirements. Where legal requirements are more stringent than these standards, legal requirements prevail.
2. Definitions
- Capability Disclosure
- A publicly accessible statement describing the intended function and scope of an AI system's outputs.
- Limitation Disclosure
- A statement identifying material constraints on an AI system's reliability, accuracy, or applicability, including known conditions under which performance degrades.
- Training Data Provenance
- A summary description of the data sources used to train the AI system, at a level of detail sufficient to allow affected persons to assess potential biases or data coverage gaps.
- Decision Explainability
- The ability to provide a human-readable account of the principal factors contributing to a specific AI system output upon request.
3. Requirements
4. Assessment Criteria
Assessment is conducted through review of published disclosure documents, user interface inspection to verify notification practices, and testing of explainability mechanisms where applicable. Assessors will verify that disclosures are accurate, current, and accessible without barriers.
5. Compliance Indicators
| Requirement | Compliant Indicator | Non-Compliant Indicator |
|---|---|---|
| REQ-002-1 | Capability disclosure publicly accessible, current, and accurate | No disclosure, or disclosure requires authentication to access |
| REQ-002-2 | Limitation disclosure current and updated within required intervals | No disclosure, or known limitations not reflected in published disclosure |
| REQ-002-3 | Training data summary published at category level or greater | No summary, or summary contains material inaccuracies |
| REQ-002-4 | Explainability mechanism exists and produces accessible outputs | No mechanism for requesting explanation, or mechanism non-functional |
| REQ-002-5 | AI involvement notification present at point of output delivery | No notification, or notification requires user action to discover |
6. Version History
| Version | Date | Notes |
|---|---|---|
| v1.1 | 2026-02-01 | REQ-002-5 amended to clarify notification timing requirement. Minor editorial corrections. |
| v1.0 | 2026-01-15 | Initial publication. Adopted by Standards Board resolution 2026-01-12. |