Compliance

Provenance

Record of a content item's origin and transformation history.

How the laws define this

Both instruments agree that provenance consists of metadata and traceability mechanisms that document the creation and processing steps of generated content to verify authenticity and enable accountability. Where they diverge is in the exact links and tools required to establish this lineage. The Deepfakes Guidelines (SDAIA) focus on tracking content through its origin, creation, and subsequent transformations using cryptographic markers and metadata, aiming to enable cross-platform tracing and machine-readable disclosure. By contrast, the Model AI Governance Framework for Generative AI extends traceability deeper into the development pipeline, requiring generated content to be linked back to specific datasets, model versions, and training sources. Consequently, while the Deepfakes Guidelines (SDAIA) emphasize content-level markers to trace material across platforms, the Model AI Governance Framework for Generative AI stresses system-level lineage to support forensic analysis and attribution.

Synthesised from the 2 statutory definitions below. Each one is quoted in full, with its source.

What it triggers

In the Saudi Arabia Deepfakes Guidelines (SDAIA), structured information documenting the origin, creation process, and subsequent transformations applied to a piece of content is used to verify authenticity, support machine-readable disclosure, and enable cross-platform tracing and accountability.source
In the Model AI Governance Framework for Generative AI, metadata, lineage information and traceability mechanisms that link generated content back to datasets, model versions, training sources and processing steps support attribution, forensic analysis, verification and accountability.

Definitions (2)

Structured information documenting the origin, creation process, and subsequent transformations applied to a piece of content (including metadata and cryptographic markers), used to verify authenticity, support machine-readable disclosure, and enable cross-platform tracing and accountability.

Deepfakes Guidelines (SDAIA)Definition 1 of 2

Metadata, lineage information and traceability mechanisms that link generated content back to datasets, model versions, training sources and processing steps to support attribution, forensic analysis, verification and accountability.

Model AI Governance Framework for Generative AIDefinition 2 of 2

Not to be confused with

content provenance information

Content provenance information refers specifically to machine-readable data documenting a content item's origin and history, while provenance encompasses broader structured information including metadata and cryptographic markers.

Content Provenance

Content provenance describes the traceability of content origin and processing history, whereas provenance includes detailed documentation of the creation process and transformations for verification purposes.

Questions and answers

What is provenance used for in the Saudi Arabia Deepfakes Guidelines?

Provenance is structured information documenting the origin, creation process, and transformations of content, including metadata and cryptographic markers. It is used to verify authenticity, support machine-readable disclosure, and enable cross-platform tracing and accountability.

How does the Model AI Governance Framework define provenance?

It defines provenance as metadata, lineage information and traceability mechanisms linking generated content to datasets, model versions, training sources and processing steps. This supports attribution, forensic analysis, verification and accountability.

Is provenance binding under these instruments?

No, the Saudi Arabia Deepfakes Guidelines (SDAIA) are a draft guideline that is not binding. The Model AI Governance Framework is a model framework with no recorded jurisdiction or binding status.