What "signer not trusted" or "signer trust not established" means in a C2PA check
It means the signature on a Content Credentials record is intact, but the certificate that made it is not on any trust list the checker loads. It does not say the signer did anything wrong, or that the claims inside are true or false. A record from a named company and a forged one can read the same.
What does "signer trust not established" mean?
Our checker shows the headline "Content Credentials found; signer trust not established" when two things are true. The file carries a signed record whose signature and match to the image hold, and the certificate that signed it is on neither of the two trust lists we load. The detail under the headline says: "It is on neither trust list, so that name is not verified; anyone can make a certificate with any name."
Other tools word this differently. The C2PA specification calls it signingCredential.untrusted.
What does the C2PA specification say about an untrusted signer?
We read C2PA Specification 2.2 on October 30, 2026.
- The code. Section 15.2.2.3 defines the code as "The signing credential is not listed on any of the validator’s applicable trust lists."
- The test. Under section 15.7, a validator checks for a chain of trust from the signing certificate to an entry on an applicable trust anchor list. If it cannot, it uses that code; if it can,
signingCredential.trusted. - Valid is not Trusted. Section 14.3 describes three states for a record: well-formed, valid and trusted. A trusted record is also valid, but a valid record, with a sound signature and no change since signing, can still be untrusted.
- The lists belong to the validator. Section 14.4.1 says a validator keeps its own set of trust anchors. That set must include the C2PA Trust List and may include more. "Untrusted" therefore means not on the lists this validator uses.
The specification words an untrusted signer as a rejected claim. Our checker separates the cases. A changed image or damaged record reads "Content Credentials failed validation". An intact signature from an unlisted signer reads "signer trust not established".
Which trust lists does our checker load?
The Content Authenticity Initiative's trust list page says conforming validator products must refer to the C2PA Trust List. It also says the older Interim Trust List was frozen on January 1, 2026, and that during the transition period a validator may consult both but must tell them apart. We do both. Our Privacy and method page describes how we use them.
| List and file | Date of our copy | Certificates |
|---|---|---|
| C2PA Trust List, signer anchors | October 17, 2026 | 32 |
| Interim Trust List, anchors | January 8, 2026 | 27 |
| Interim Trust List, allowed certificates | April 21, 2026 | 115 |
A signer added to a list after the date of our copy is not recognized until we fetch the list again.
Why did some of the ChatGPT files we tested read "signer trust not established"?
OpenAI's help article on provenance signals says images made with ChatGPT, Codex or the OpenAI API include Content Credentials. On October 30, 2026 we read the subject and issuer of every certificate in our copies of the C2PA Trust List (signer anchors and time-stamping list) and the Interim Trust List (anchors and allowed certificates). None names OpenAI. The newest sample's certificate was issued by an authority named Trufo, which is on the C2PA Trust List, so that file reads as trusted. The two older samples' certificates had expired.
| File | Name in the certificate | What the checker reported |
|---|---|---|
| Sample image from the c2pa-rs project | C2PA Test Signing Cert | Signature valid, image matches, on neither list |
| ChatGPT sample file from our research set | OpenAI | On neither list; certificate expired |
| Wikimedia Commons file filed under "Images generated by ChatGPT" | OpenAI | On neither list; certificate expired |
| A picture we drew and signed with a self-made certificate | OpenAI OpCo, LLC | On neither list; it carries the same name as the newest ChatGPT sample, which reads as trusted |
Our measured result: 3 of the 32 real files we ran read "Content Credentials found; signer trust not established". They are the C2PA test image and two ChatGPT files. The drawn picture gets the same headline, and a line saying the result cannot tell it from a genuine record. The name "OpenAI OpCo, LLC" is on both it and the trusted sample, so the name does not decide the result; the list does. All counts are on our accuracy page.
Nine of the 32 files carry a record. One read as trusted on the current C2PA Trust List. Five matched only the frozen Interim Trust List; the three real files above matched neither. Files filed under a product's name show only what someone uploaded, so we draw no conclusion about the product.
Does it mean the file is altered or the signer did something wrong?
No. Three separate questions get answered in a check, and this result answers only the third.
- Is the signature sound? In all three real files the checker found it intact.
- Does the record match the image? For all three, the result says "The image matches this record."
- Does a list we load vouch for the signer? For these three, no.
The result passes no judgment on the company named in the certificate. It reports that the lists we load do not cover that certificate. Even when a signer is trusted, the C2PA explainer says Content Credentials do not give a verdict on whether the contents are true. It adds that trust rests on the ecosystem where the credential is used, and that a person may keep a personal trust list.
One limit applies. Our checker does not go online to ask whether a certificate was revoked, because that would send a request from your browser to a third party.
What this means when you check an image
Open the checker, choose the file, and look for the headline "Content Credentials found; signer trust not established". It tells you a signed record is present, and that its signature and match to the image were verified. It shows the name in the certificate as "The signing certificate names" that name, never as "Signed by".
It does not tell you who really signed, whether the record's claims are true, or how the image was made. The findings under it carry the tier "Signed, signer not trusted"; read them as statements by whoever holds that certificate.
If the file claims to come from a company, ask that company's own tool. OpenAI runs a verification page for files from its tools. It opens another company's site; you may need to sign in, and your image goes to them, not to us. OpenAI says its tool looks for supported signals, such as a SynthID watermark or a trusted C2PA manifest associated with OpenAI. We do not check SynthID, and we cannot see how OpenAI decides trust.
Guides on this topic
- What are Content Credentials? is the main page on this topic.
- The C2PA Trust List and the Interim Trust List shows who is on which list.
- Can Content Credentials be faked? covers self-made certificates.
- Does ChatGPT add a watermark or metadata? covers what OpenAI says it adds.
- How to check Content Credentials walks through the other results.
Frequently asked questions
- Is "signer not trusted" the same as "failed validation"?
- No. "Failed validation" means the record is damaged or the image changed after signing. "Signer trust not established" means the signature and the match are intact and no list we load covers the signer.
- Why do two lists matter?
- The Interim Trust List was frozen on January 1, 2026. A signer on only one list gets a different headline, so we show which list matched.
- Can the result change later?
- Yes. If a list adds the signer's certificate and we load the new copy, the same file would read differently.