Home / Standards / Content Integrity

Content Integrity Standard

Standard ID: GCAIS-STD-004
Standard IDGCAIS-STD-004
Versionv1.0
Published2026-03-01
StatusActive
Review Due2028-03-01 (24-month cycle)
SupersedesInitial publication

Table of Contents

  1. Scope
  2. Definitions
  3. Requirements
  4. Assessment Criteria
  5. Compliance Indicators
  6. Version History

1. Scope

This standard establishes requirements for organizations operating AI systems that generate, substantially modify, or distribute media content, including text, images, audio, and video. It applies to systems where AI-generated or AI-modified content is presented to end users or distributed through public or semi-public channels.

The requirements are intended to ensure that the origins and modification history of AI-processed content are appropriately disclosed and technically traceable. They apply to the organization operating the AI system and to content produced by that system, whether or not the content is published under the organization's own name.

This standard does not apply to AI systems used solely for internal processing tasks where outputs are not presented to end users or distributed externally.

2. Definitions

AI-Generated Content
Media content produced substantially by an AI system, including text, images, audio, video, and composite media, where the AI system's contribution is material to the content's form or substance.
Provenance Metadata
Structured data embedded in or associated with a content item that records its origin, creation process, and modification history.
Synthetic Media
Media content generated or substantially modified by AI to depict persons, events, or conditions that did not occur, or to alter the representation of persons or events in a material way.
Watermark
A detectable or imperceptible signal embedded in media content that identifies it as AI-generated or AI-modified, survives reasonable processing or compression, and is detectable by the issuing organization's tools.

3. Requirements

REQ-004-1: AI-Origin Labeling The organization shall ensure that AI-generated content distributed to end users is labeled at the point of delivery as having been generated or substantially modified by an AI system. Labels shall be clear, prominent, and shall not require users to take additional steps to view.
REQ-004-2: Provenance Metadata The organization shall implement a mechanism to attach provenance metadata to AI-generated content at the point of creation. Metadata shall identify at minimum: the generating system, the date and time of generation, and a unique content identifier. Metadata standards used shall be documented.
REQ-004-3: Manipulation Disclosure Where AI systems are used to materially alter content depicting real persons, events, or conditions, the organization shall ensure disclosure of the nature of the alteration accompanies the content wherever it is distributed. Disclosure shall be specific enough to allow a reasonable person to assess the extent of AI modification.
REQ-004-4: Synthetic Media Watermarking AI systems producing synthetic media shall embed a detectable watermark in all outputs at the point of generation. The watermark standard used shall be documented. The organization shall maintain tools capable of detecting watermarks in content it has generated.
REQ-004-5: Removal Mechanisms for Unlawful Content The organization shall maintain and document a mechanism by which AI-generated content that is identified as unlawful or in material violation of the organization's content policies may be removed or disabled within a defined response window. The mechanism shall be tested at intervals not exceeding 12 months.

4. Assessment Criteria

Assessment includes review of labeling implementation through sampling of distributed content, inspection of provenance metadata implementation, review of watermarking documentation and detection tool testing, and verification of removal mechanism response records.

5. Compliance Indicators

Requirement Compliant Indicator Non-Compliant Indicator
REQ-004-1 AI-origin label present at point of delivery on sampled content No label, or label requires user action to discover
REQ-004-2 Provenance metadata present and recoverable on sampled content No metadata, or metadata missing required fields
REQ-004-3 Manipulation disclosure accompanies altered content in distribution No disclosure, or disclosure absent from distribution channels
REQ-004-4 Watermarks detectable in sampled synthetic media; detection tool functional No watermarking, or watermarks not detectable by organization's tools
REQ-004-5 Removal mechanism documented, tested, and response records retained No mechanism, or mechanism not tested within required interval

6. Version History

Version Date Notes
v1.0 2026-03-01 Initial publication. Adopted by Standards Board resolution 2026-02-26.
GCAIS-STD-004 Content Integrity Standard v1.0 (HTML) Request PDF version