Use Case

Cryptographic Provenance for
AI-Retrieved Content

When AI systems retrieve document chunks for generating responses, there is no way to verify the source's integrity. Diogenes brings verifiable provenance to RAG pipelines, enabling citations with cryptographic proof.

The Problem

AI Citations Without Provenance Are Meaningless

Retrieval-Augmented Generation (RAG) has become the standard pattern for grounding AI responses in real documents. But the documents themselves have no verifiable provenance. An AI that cites a "source" cannot prove that source is authentic, unaltered, or authored by who it claims.

  • Retrieved document chunks lack any proof of source integrity or authorship.
  • Poisoned or tampered documents in the knowledge base propagate as trusted citations.
  • Users cannot independently verify the provenance of AI-cited sources.

RAG Pipeline with Diogenes

Documents are signed at ingest
Each document carries its attestation graph into the vector store
AI retrieves chunks
Retrieved chunks carry manifest references
Citations include provenance
Users can verify the source independently
Trust assessment per source
Evaluate each cited source against your trust network
How Diogenes Helps

Cryptographic Provenance for Every Retrieved Chunk

Manifest-Linked Chunks

Documents ingested into RAG pipelines carry verifiable attestation graphs. Each chunk traces back to its signed manifest, so the connection between the AI's citation and the original source is cryptographically provable.

Tamper Detection

If a document in the knowledge base is altered after signing, the hash mismatch is immediately detectable. Poisoned documents cannot masquerade as verified sources.

Trust-Aware Retrieval

RAG systems can prioritize sources from signers within the user's trust network, or flag citations from untrusted sources. Trust becomes a first-class signal in retrieval ranking.

More Than a Signature

How Diogenes Compares

Existing approaches to AI content trustworthiness rely on watermarks, metadata, or blind trust in the knowledge base. Diogenes provides cryptographic proof that connects every retrieved chunk to a verified human author.

Capability C2PA / Content Credentials RAG Without Provenance Diogenes
Chunk-level verification File-level only None Every chunk traces to its signed manifest
Author identity CA-issued certificates Filename/URL attribution Endorsement-verified cryptographic identity
Trust assessment Binary (signed or not) None Trust score based on your network and policies
Tamper detection Metadata can be stripped None Hash mismatch immediately detectable
Retrieval ranking Not applicable Relevance only Trust-aware: prioritize verified sources

Ready to bring verified provenance to your AI pipeline?

Read the Documentation Try Verification