Latent disclosure
Embedded, machine-readable provenance metadata.
How the laws define this
Both statutory formulations agree that latent disclosures consist of non-visible or invisible, machine-readable data embedded directly inside AI-generated content to enable the verification of its origin. Each description emphasizes that this hidden data communicates basic details identifying the output, specifically citing the provider's name and the underlying system version. The primary distinction between the two instruments lies in how they frame the disclosure's durability, technical scope, and specific data points. Under California SB 942 — California AI Transparency Act, the definition focuses on metadata designed to persist through sharing and common transformations to facilitate programmatic verification of synthetic origin, explicitly referencing timestamps and unique IDs. In contrast, the formulation under the California AI Transparency Act defines the disclosure as permanent provenance data intended to verify content authenticity and origin, citing a creation date instead of timestamps or unique IDs.
Synthesised from the 2 statutory definitions below. Each one is quoted in full, with its source.
How jurisdictions differ
What is distinctive in each definition, binding law in force first.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| California, United States | California SB 942 — California AI Transparency Act | In Force (Amended) | Specifies examples of embedded data (provider name, system/version, timestamp, unique ID) and emphasizes persistence through sharing and common transformations. |
| California, United States | California AI Transparency Act | In Force (Amended) | Highlights permanence of the data and focuses on conveying provider name, system version, and creation date for authenticity verification. |
| California, United States | California SB 1000 — California AI Transparency Act (2026 Amendments) | In Force | Broadens to generated or altered media and explicitly lists required elements including whether content was created or altered by GenAI. |
What it triggers
- In California, under SB 942, large generative AI providers must offer free AI detection tools and embed content provenance disclosures.
- In California, under the AI Transparency Act (SB 942, amended by AB 853), generative AI providers and platforms must embed disclosures in synthetic content.source
- In California, under SB 1000, expansions require provenance disclosures, privacy safeguards, and enforcement mechanisms.source
Definitions (4)
An invisible, machine-readable metadata disclosure embedded in AI-generated content (e.g., provider name, system/version, timestamp, unique ID) designed to persist through sharing and common transformations to enable programmatic verification of synthetic origin.
Latent disclosures are machine-readable and permanent provenance data embedded within AI-generated content. This data is not immediately visible but is designed to convey information such as the provider's name, system version, and creation date, enabling verification of content authenticity and origin.
Latent disclosure is permanent or extraordinarily difficult-to-remove provenance data embedded in generated or altered media. It must identify the provider, system name and version, creation or alteration time, a unique identifier, and whether the content was created or altered by the GenAI system.
A disclosure included in content under the chapter's applicable provisions. Provider disclosures convey specified origin and modification information and, to the extent technically feasible, meet durability, verification compatibility, and interoperability requirements.
Not to be confused with
Manifest disclosure refers to visible, human-readable labeling, while latent disclosure involves invisible, machine-readable metadata.
Labelling/watermarking encompasses both visible and embedded markers, whereas latent disclosure specifically means embedded, machine-readable provenance metadata.
Questions and answers
What must latent disclosures contain under California law?
They must include the provider name, system/version, timestamp, unique ID, or creation date, and indicate if content was created or altered by GenAI. The data is designed to persist and enable verification.
Which AI providers are required to use latent disclosures in California?
Large generative AI providers must embed these disclosures in AI-generated content under SB 942 and the AI Transparency Act.
When did California's latent disclosure requirements take effect?
The requirements are in force as of January 1, 2026.
Related Terms
Content Provenance
Traceability of content origin and processing history....
Manifest disclosure
Visible, human-readable AI content labeling....
AI Attribution
Obligation to identify and cite the AI system behind generated content....
Provenance data
Data used to trace and verify content origin....
Labelling / Watermarking
Visible or embedded markers indicating synthetic origin....