RAG Source Attestation
Bind each AI citation to a ZK proof of the exact document version (docHash) it claims to reference, so citation integrity holds cryptographically even after the index is rebuilt.
Three voices from the front line.
- IT / digital transformation
“We have no way to confirm a RAG answer's citation is really the latest, authentic version”
- AI engineering
“We can't detect an untrusted source slipping into the RAG pipeline”
- Audit / quality
“We need to reconstruct and explain where an AI answer came from at audit time”
Hand over the source, or just the facts?
Nothing changes on the floor. Everything changes for the receiver.
① Your team just saves, as always.
- The usual step
- Fill in the record and save
- On save
- A proof is attached (API, behind the scenes)
- The document itself
- never sent
② They just open a link.
- Proven fact
- each citation is bound to its source version
- the model's internal context
- not shown
- Login / keys
- not needed
- The document stays private — the record itself is never sent or disclosed.
- Independent verification — the receiver just opens a link. No account, no keys.
- Edits are detected — even a one-character edit fails verification.
Each source the AI cites is cryptographically bound to the exact document version that citation points to. A citation becomes a verifiable reference, not just a label. The original body reaches neither the index nor the answer; what leaves is only the fact that "this citation comes from the version valid at answer time." Even if the vector DB is rebuilt or the policy is revised, the citation proof attached to a past answer stays immutable.
See the technical details ↗Why the usual methods fall short.
Only work that needs all three at once — pass without exposing, independent verification, tamper-evidence — is Lemma's domain.
| Method | Pass without exposing | Independent verification | Tamper-evident | What happens |
|---|---|---|---|---|
| Access control / permissions | △ | ✗ | ✗ | “Someone inside could have edited it” remains possible |
| Masking / redacted copies | △ | ✗ | ✗ | Redaction work grows; the original is still unproven |
| Encrypt and store / send | ✓ | ✗ | ✗ | To verify, the receiver needs it disclosed after all |
| Lemma (ZK proof)the only one with all 3 | ✓ | ✓ | ✓ | The receiver just opens a link |
How it works — and how to start.
We help design disclosure scope and retention, run the PoC, and support production.
Start with a 30-minute call.
Tell us one AI workflow where "citations appear, but you can't prove they're real" applies, in the first 30 minutes. No answer logs or source documents required.
Related Use Cases
Find the case that matches your industry and problem
Browse all 36 use cases →TRY LEMMA
Run it yourself.
No sales call needed — start hands-on with Lemma's products.