AI Media
An informational, non-legal guide to provenance, Content Credentials, C2PA, manifests, signatures, source ingredients, licenses, attribution, usage rights, consent, AI disclosure, and training preferences.
17 matching terms
Provenance
Provenance
Information about a media asset's origin, creation process, edits, and movement through a production workflow.
Record it so reviewers and audiences can understand where an asset came from and how it changed.
record source file, creator or provider, creation date, tools used, edits, and publication versionContent Credentials
Content Credentials
Tamper-evident provenance information cryptographically bound to digital content and presented as a record of how it was created or edited.
Use supported tools to attach or inspect creation, edit, tool, and AI-use information before publication.
Content Credentials can support transparency, but they do not by themselves prove that depicted events are true or that every right is cleared.
attach Content Credentials that identify the creation tool, edit history, and AI-assisted stepsContent Credentials
C2PA
The Coalition for Content Provenance and Authenticity and its open technical standard for verifiable media provenance.
Use the term when checking whether tools, platforms, and files can create, preserve, or verify compatible credentials.
verify that the export and publishing workflow preserves C2PA-compatible credentialsContent Credentials
C2PA manifest
A signed data structure containing assertions and related information about an asset's provenance.
Inspect it when validating which organization or tool signed the credential and what information it contains.
inspect the active manifest, signer, assertions, validation status, and ingredient referencesContent Credentials
Assertion
A structured statement inside a Content Credential about the asset, its actions, metadata, or AI involvement.
Review individual assertions instead of treating the credential badge as one undifferentiated claim.
list assertions for creation, edits, ingredients, software actions, and generative AI useContent Credentials
Cryptographic signature
A mathematical signature used to verify who or what signed provenance data and whether it changed afterward.
Validate it to detect tampering and to evaluate the signer through the applicable trust system.
validate the credential signature, certificate chain, signer, and timestamp before publicationProvenance
Ingredient
A source asset incorporated into another asset, potentially carrying its own provenance information.
Track significant source images, clips, audio, or documents used in a composite or generated result.
record each significant ingredient with source URL, license, creator, and credential referenceContent Credentials
Soft binding
A durable linking method such as watermarking or fingerprint lookup that can help rediscover provenance data when embedded metadata is removed.
Use it when credentials must remain discoverable across metadata-stripping distribution channels.
No binding method guarantees that provenance will survive every transformation or distribution path.
enable a supported soft binding and verify credential recovery after platform uploadRights and licenses
Rights holder
A person or organization that owns or controls one or more rights relevant to using a media asset.
Identify the relevant rights holder before requesting permission, licensing, publishing, or sublicensing.
record the rights holder, contact, covered asset, granted uses, territory, and termRights and licenses
License
A permission framework that states how an asset may be copied, modified, distributed, displayed, or used commercially.
Check the exact license name, version, scope, restrictions, attribution, and expiration rather than assuming online availability means free use.
License interpretation and applicable rights vary by jurisdiction and project; obtain qualified advice when the risk is material.
verify license name and version, commercial use, derivatives, territory, term, and attribution requirementsRights and licenses
Attribution
Credit that identifies the creator, source, license, and often whether changes were made.
Provide it in the placement and form required by the license or permission record.
credit: title, creator, source link, license name and link, and note of modificationsRights and licenses
Usage rights
The specific permitted uses of an asset, which may be limited by medium, audience, territory, duration, modification, or commercial context.
Record rights at the project and asset level so exports are not published beyond approved scope.
approved for paid social video in Korea for six months; editing allowed; sublicensing not allowedConsent and releases
Model or property release
A documented permission related to using a recognizable person or certain private property in specified media and contexts.
Check whether the planned generation, alteration, advertising, distribution, and duration fall within the release.
Release requirements differ by jurisdiction, subject, property, and use.
confirm the signed release covers AI-assisted edits, commercial publication, channels, territory, and termConsent and releases
Likeness consent
Informed permission to capture, generate, edit, or publish an identifiable person's face, body, or recognizable appearance.
Define the intended synthetic changes, sensitive contexts, audience, channels, and withdrawal process before production.
document consent for the specified likeness edit, campaign context, distribution channels, and durationConsent and releases
Voice consent
Informed permission to record, synthesize, clone, transform, or publish an identifiable person's voice.
Keep the approved purpose, script types, languages, model access, retention, and revocation terms with the voice asset.
record approved voice uses, languages, model access, storage period, and revocation contactDisclosure and preferences
AI-use disclosure
A clear notice that AI generated or materially altered part of a media asset.
Use it where audiences, clients, platforms, contracts, or applicable rules require transparency.
Disclosure requirements depend on platform, contract, location, content type, and context.
disclose that the background and narration were AI-generated and the final edit was human-reviewedDisclosure and preferences
Training and use preference
Metadata or a service setting that communicates whether a creator wants content used for generative AI training or related purposes.
Apply supported preferences consistently and retain them in the asset and publishing workflow where possible.
A preference signal is not a universal technical or legal enforcement mechanism; support and effect vary by system.
set the supported preference to prohibit generative AI training and verify it remains attached after exportOfficial material referenced when preparing the technical terminology on this page.
C2PA-related technical terms on this page were independently summarized with reference to the official specification.
C2PA Technical Specification 2.4Creative Commons Attribution 4.0 International (CC BY 4.0)