Content Credentials
Content Credentials is an open, royalty-free standard that attaches tamper-evident, cryptographically signed provenance to images, video, audio and documents. It is stewarded by the C2PA, a Joint Development Foundation project backed by more than 500 companies.
What is Content Credentials?
Content Credentials is an open technical specification for the provenance of digital content, described on the site as a nutrition label for digital media. It is produced by the Coalition for Content Provenance and Authenticity (C2PA), a project of the Joint Development Foundation, a Washington-based 501c6 non-profit. The coalition was founded on 22 February 2021 by Adobe, Arm, BBC, Intel, Microsoft and Truepic, merging the efforts of the Content Authenticity Initiative and Project Origin. Its steering committee now brings together Adobe, Amazon, BBC, Google, Meta, Microsoft, OpenAI, Publicis Groupe, Sony, TikTok and Truepic, alongside more than 500 companies spread across three membership tiers.
Technically, a Content Credential — also called a C2PA Manifest — holds cryptographic hashes of the asset together with its provenance data. Any alteration breaks that link, which is what makes the record tamper-evident. The foundations are ordinary and well tested: SHA-256 hashes, X.509 certificates and digital signatures. Signing certificates are issued by certificate authorities listed on the C2PA Trust List, and each certificate carries an assurance level. Existing metadata standards — IPTC, XMP and EXIF — can be wrapped as tamper-evident assertions, so the format sits alongside what photographers and newsrooms already use.
Three mechanisms work together. Embedded provenance attaches the signed data to the content. Invisible watermarking and digital fingerprinting act as soft bindings, letting a credential be found again once metadata has been stripped. Photos, video, audio files and documents are all covered, and signing works offline: a camera can sign with locally stored keys, no connection required. The size overhead is modest, on the order of a kilobyte per asset.
The official Content Credentials icon, unveiled on 10 October 2023 after two years of research and design and presented as an icon of transparency, marks a file that carries provenance information and opens an interactive panel showing it.
The specification is released under a royalty-free licence: no licence fee, and no obligation to join the organisation in order to implement it. The 1.x series is treated as legacy, while the 2.x series — 2.1, then 2.3 — is the recommended one.
What it does
- Attach a signed record of origin to an image, a video, an audio file or a document
- Add a fresh signature at every edit, recording the editing method, the place, the date and the time
- Open the Content Credentials icon to read a file's provenance log
- Check whether a file is authentic through the online Verify tool
- Signal that a piece of content was generated or retouched by an AI model
- Recover a lost provenance record through invisible watermarking or digital fingerprinting
- Submit a generator product for assessment under the C2PA conformance programme
When to use Content Credentials / When not to
A quick filter to help you decide if Content Credentials is the right fit.
When to use Content Credentials
- Photographers and journalists who need to prove where their images and footage came from
- Newsrooms and publishers that have to document the full editing history of a media file
- Software vendors and device manufacturers that want to sign the content their products create
- Platforms that want to show their users where a piece of content originated
- Generative AI teams that must disclose when content has been created or retouched by a model
When not to use Content Credentials
- Anyone shopping for digital rights management: Content Credentials never restricts access to or use of a file
- Teams hoping for a deepfake detector: the standard reads a signed, declared provenance instead of analysing the content itself
- Users who need to tie a file to a named author: the core specification deliberately avoids attributing content to individuals or organisations
- Anyone expecting missing credentials to prove a forgery: a great deal of authentic content carries none at all
- Investigators who need a fully verified chain: without access to the ingredient data, only a prior validation can be confirmed
How to use Content Credentials
A typical end-to-end flow, from setup to results.
- Spot the Content Credentials icon on a piece of media and open it to read the history, from origin through to the version in front of you
- Drop the file on the online Verify tool, linked from the site menu, to check its credentials for yourself
- Consult the glossary on the site if the vocabulary of provenance is new to you
- As a creator or software vendor, connect provenance information to the content your tool produces
- Sign again at every update, so that the editing method, the place, the date and the time are added to the credential
- As an implementer, read the specification, choosing the recommended 2.x series over the legacy 1.x one
- Test your implementation against the public test files published by the C2PA on GitHub
- Express interest in the conformance programme through the form linked from the conformance page
- Pass the conformance assessment, obtain a Claim Signing Certificate from a Trust List certificate authority, and get listed on the Conforming Products List
- Browse the live trust and product lists in the C2PA Conformance Explorer, and join the open source community on the Content Authenticity Initiative Discord
Pros & Cons
Pros
- Open, royalty-free standard: implementable with no membership and no licence fee
- Adopted by the largest platforms and vendors, from Adobe, Google and Microsoft to OpenAI, Meta, Amazon, TikTok, Sony and the BBC
- Proven cryptographic foundations — SHA-256, X.509, digital signatures — that make any alteration detectable
- Durable by design: watermarking and fingerprinting can restore provenance after metadata has been stripped
- Interoperable with existing IPTC, XMP and EXIF metadata, and usable offline, including inside a camera
- Public governance: certificate policy, security requirements and conformance lists are all published on GitHub
- Privacy-preserving and lightweight: no mandatory attribution to a person, and roughly a kilobyte added per asset
Cons
- Metadata can be stripped, deliberately or not; soft bindings soften the loss without guaranteeing recovery
- Missing credentials prove nothing, so coverage depends entirely on adoption by tools and platforms
- Certificate fees are set by each certificate authority and are not published by the C2PA
- Conformance assessment and legal onboarding are heavy prerequisites for reaching the Conforming Products List
- 1.x implementations are declared legacy, and the Interim Trust List was frozen on 1 January 2026, leaving existing certificates to expire
- The core specification does not attribute content to an identified author; attribution relies on community extensions
- Nothing on the organisation's side about the GDPR, no data processing agreement, no sub-processor list, and no hosted API or SDK in the c2pa-org GitHub organisation
Pricing & Plans
Content Credentials is free of charge. The specification is released as an open standard under a royalty-free licence, so no licence fee applies, and there is no paid plan, no subscription and no trial to speak of; C2PA members have undertaken to make their contributions freely available on those terms, and membership is not required in order to implement the standard. Two optional costs sit outside the specification itself. An organisation seeking formal recognition must obtain a Claim Signing Certificate from a certificate authority on the C2PA Trust List, and those fees are set by each authority under its own commercial terms rather than published by the C2PA. Joining the coalition is likewise possible, through an application form and a membership agreement, but no membership fee is disclosed on the site.
- the specification is free for everyone
- and the tiers below govern participation in the coalition
- not access to the standard
- eleven organisations — Adobe
- Amazon
- BBC
- Meta
- Microsoft
- OpenAI
- Publicis Groupe
- around thirty-eight organisations
- among them Ad-ID
- Arm
- Bloomberg
- Canon
- Deloitte
- DigiCert
- ElevenLabs
- several hundred organisations
- No price is attached to any of these tiers on the site
Data, GDPR & hosting
A consolidated view of how Content Credentials handles your data.
GDPR overview
There is no mention of the GDPR anywhere. A search across the 34 pages collected — contentcredentials.org, c2pa.org and the Joint Development Foundation policy that governs both — returns zero occurrences of GDPR or General Data Protection Regulation. No Article 27 EU representative is named, no data protection officer is appointed, no data processing agreement is offered, no sub-processor list is published and no data hosting country or region is declared. The governing law is stated as that of the State of Washington, United States, and the legal entity is registered in Wilmington, Delaware. The privacy policy does set out seven general data protection principles, but attaches them to no European instrument. European readers should weigh that against the nature of the tool: the standard involves no account and collects nothing, so the gap concerns the organisation's website, not the technology.
Who owns the data?
There is no user account and no hosted service on contentcredentials.org: the site is informational. Content Credentials are attached to the file itself rather than stored with a third party, so a signed asset and its provenance record stay with whoever holds the file, and the site makes no claim over user content. Personal data collected through the site falls under the Joint Development Foundation's generic policy, shared by all its projects, which states that it will “collect and use of personal information solely with the objective of fulfilling those purposes specified by us and for other compatible purposes, unless we obtain the consent of the individual concerned or as required by law.”
Reuse rights
The specification itself is published as an open standard under a royalty-free licence, so anyone may read it, implement it and build on it without asking permission and without paying a fee; C2PA members have undertaken to make their contributions freely available on those terms. For personal data, the Joint Development Foundation policy sets out general principles rather than reuse rights: purposes are identified “before or at the time of collecting personal information”, data is gathered by lawful and fair means with the knowledge or consent of the individual where appropriate, kept relevant, accurate, complete and up to date, and protected by reasonable security safeguards against loss, theft, unauthorised access, disclosure, copying, use or modification. No advertising use, no resale and no training of models on customer data is mentioned anywhere.
Data retention & training
Hosting summary
No hosting country or region is declared for user data anywhere on the site — and here that is mostly because there is no user data to host. Content Credentials are embedded in the file itself, so the model needs no central database: a signed asset carries its own provenance wherever it travels. Cloud retrieval is mentioned as a complementary mechanism for finding a credential again, without naming any host. What can be established concerns the website rather than the technology. A network probe places contentcredentials.org behind Fastly (AS54113) at 23.185.0.4, an anycast node resolving to San Francisco, United States. The legal entity, Joint Development Foundation Projects, LLC, is registered in Wilmington, Delaware, United States, and the terms name the law of the State of Washington as governing. For a European organisation that means an American jurisdiction, with no declared EU hosting, no data processing agreement and no sub-processor list — a gap on the organisation's side, not in the standard.
Things to keep in mind
Risks and trade-offs to weigh before adopting Content Credentials.
- Treating the presence of a credential as proof that the content is true: it records origin and edits, never truthfulness
- Reading missing credentials as evidence of a fake — most authentic media still carries none, and the reflex breeds unwarranted suspicion
- Using it as proof of identity: the core specification does not attribute content to a named author, and community extensions are not the same thing
- Assuming the chain is complete: without the ingredient data, only a prior validation can be confirmed, not the whole history
- Relying on metadata that a single upload can strip; soft binding recovery is a mitigation, not a guarantee
- Overlooking the 2026 transition: the Interim Trust List was frozen on 1 January 2026, 1.x implementations are legacy, and certificate costs are set, unpublished, by each authority
- Expecting privacy paperwork that does not exist: the applicable policy is the generic Joint Development Foundation one, with no GDPR mention, no DPA, no sub-processor list, and the site publishes no email address of its own
Setup & Integrations
Technical difficulty
Difficulty depends entirely on your role. To consult a credential, nothing is required: the icon and the online Verify tool are enough, with no account to create. To implement the standard, expect real engineering — reading the 2.x specification, handling X.509 certificates, signing cryptographically. To be recognised as conforming, add a technical and security assessment, legal onboarding and a certificate from a Trust List authority, with hardware-backed attestation sometimes required at enrolment. Public test files, the Certificate Policy, the Generator Product Security Requirements, the Conformance Explorer and a community Discord support that work.
Deployment
Integrations
Behind Content Credentials
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Is Content Credentials free to use?
Do I have to join the C2PA to implement the specification?
Is this a form of DRM?
What happens if the C2PA metadata is removed from a file?
Can a Content Credential be forged?
Does it work offline?
Does a Content Credential identify the author of a file?
How much does a signing certificate cost?
Are 1.x implementations still accepted?
Who governs the standard?
Should you pick Content Credentials?
Content Credentials has become the de facto reference for media provenance. Its adoption by Adobe, Google, Microsoft, OpenAI, Meta, Amazon, TikTok, Sony and the BBC, alongside more than 500 companies, gives it a reach nothing else matches, and because it is free and open there is no barrier to implementation: no licence fee, no obligation to join the coalition.
What it delivers is narrow and well defined. A signed, tamper-evident record travels with the file, saying where the content came from and how it was edited — including whether an AI was involved. It does not tell you whether an image is true, and it does not name an author: the core specification avoids attribution on purpose, to stay privacy-preserving.
Its value therefore rests on two things outside the standard. The first is adoption: a credential is only useful if the tools and platforms along the chain write it and display it. The second is metadata survival — stripping is trivial, and the soft bindings that recover a credential from a watermark or a fingerprint help without promising anything. In practice, the absence of Content Credentials tells you nothing at all.
Governance is public and documented, with the certificate policy, the security requirements and the conformance lists all published, but the ecosystem is mid-transition. The 1.x series is legacy, the Interim Trust List was frozen on 1 January 2026, and any organisation wanting formal recognition faces a conformance assessment, legal onboarding and certificate fees set by individual authorities and never published.
The honest summary is that this is not a product. It is a standard and a compliance ecosystem, free to read and free to implement, valuable to anyone who needs to show where their content came from, and worth nothing if no one along the chain looks.
- Choosing a selection results in a full page refresh.
- Opens in a new window.