V3pi produces cryptographically verifiable evidence showing whether documents or AI-generated outputs are structurally consistent or inconsistent — and precisely where differences occur.
Independent of the model layer. No probability scores. No interpretation. Just evidence.
V3pi compares two or more documents and determines whether they are structurally consistent or inconsistent. It identifies exactly which structural patterns are shared, which exist only in one document, and where those differences occur.
V3pi extracts structural representations from both documents using a proprietary methodology. These representations are compared deterministically, producing a precise inventory of what is shared, what was added, what was removed, and what changed in frequency.
The result is a conformance determination — Consistent, Inconsistent, or Review Recommended — supported by a complete evidence package that qualified professionals can examine, verify, and use.
The conformance determination is driven by an explicit verification policy included in the evidence package. The policy defines, for each structural level, whether differences must match exactly, may be flagged for review, or may be reported without affecting the determination. There are no hidden thresholds, no scoring algorithms, and no percentage-based similarity metrics. The policy itself is part of the audit trail and can be independently inspected.
V3pi does not interpret meaning, assess intent, or draw conclusions. It provides structural evidence. Professional judgment stays where it belongs — with the professionals.
Documents change. Contracts are revised between signing rounds. Regulatory notices are updated across versions. Policies are edited after board approval. AI systems generate documents that may not match their source material. Autonomous agents take actions without structural proof of conformance.
Diff tools show you which words changed, but produce visual markups — not evidence. They are productivity tools for editing, not proof of what happened.
AI-based comparison tools estimate similarity with confidence scores, but their results are probabilistic. They cannot guarantee determinism, and the tool itself may introduce errors — AI checking AI.
Legal, regulatory, and scientific environments require something different: reproducible proof of what changed, where it changed, and evidence that can be independently verified by any party.
Every engagement produces a comprehensive evidence package. All artifacts are cryptographically hashed and independently verifiable.
A high-level determination of consistency or inconsistency across the document pair, with counts of shared and divergent structural units.
Every detected structural difference, with positional anchors showing exactly where in each document the difference occurs, including surrounding context.
An inventory of all structural patterns shared between both documents — what did not change — with frequency counts and positions.
Structural patterns present in both documents but occurring at different frequencies. Reveals systematic compositional shifts that word-level comparison cannot detect.
Exact positions and character offsets for every divergent structural unit across both documents. The most granular level of evidence available.
SHA-256 hashes for every delivered file, with a complete provenance chain recording inputs, outputs, generation metadata, and the verification policy that governed the conformance determination.
A utility that any party can run — on Linux, Windows, or macOS — to independently verify the integrity of every artifact in the evidence package. No access to V3pi required.
When an AI system generates or modifies a document, V3pi verifies whether the output is structurally consistent with the reference. Because the verification layer operates independently of the model layer, the verification is independent of the system that produced the document. The verification tool itself cannot hallucinate.
Agent communication protocols handle identity and routing. V3pi handles structural verification. When an agent produces or processes a document, V3pi verifies whether the output conforms to policy — and produces witnessed evidence for every determination. V4pi, currently in beta, extends this engine to inspect agentic agents in real time, enforcing constraints at the point of execution and failing closed if actions do not conform to protocol.
A redline shows you what changed. V3pi produces an evidence package that proves what changed — with positional anchors, structural divergence reports, cryptographic hashes, and a verification utility that opposing counsel can run independently.
SHA-256 hashes on every artifact. A standalone verification utility your auditors can run independently. A complete provenance chain from input documents through final determination. Evidence designed for regulatory scrutiny.
Verify that manuscripts, protocols, and research documents have not been altered between submission versions. Deterministic results that are reproducible and independently verifiable.
V3pi operates as a scoped professional verification service. Each engagement defines the documents under review, the deliverables to be produced, and the terms of engagement.
Engagement terms, including document handling, are defined prior to processing. Evidence packages are delivered as offline artifacts with standalone verification capability.
Engagements are scoped by document set complexity. Contact us to discuss your verification needs.
V3pi's structural verification engine is not limited to regulatory text. The same deterministic approach applies wherever structured representations must be verified under explicit policy — including high-assurance workflows in AI, automation, and autonomous agent systems.
A complete V3pi evidence package is available for public inspection. This demo illustrates a policy-governed structural verification run on a real regulatory document pair.
The reference document is a section of HIPAA §164.520 — the federal regulation governing Notice of Privacy Practices. The evaluated document is a modified version with eight deliberate changes: word substitutions, negation removals, negation insertions, and logical operator changes. V3pi detected all eight structural modifications across every level of analysis.
These documents were chosen because they are public-domain regulatory text with non-trivial structure — sections, lists, obligations, and exceptions — representative of real-world compliance and legal workflows where proof of structural change matters.
Open summary.csv — one row showing the determination and counts.
Open divergence.csv — every detected structural difference with excerpts from both documents.
Open location_detail.csv — exact positions and character offsets for every divergent structure in both documents.
Open provenance.json and look under the policy key — the explicit rules that drove the conformance decision.
Run the standalone verification utility against inventory.json — confirm every file's SHA-256 hash matches.