Metastructured Data

Data that proves its own integrity.

Metastructured Data embeds cryptographic provenance, compliance metadata, and tamper-evidence directly into every AI-processed document, creating an immutable chain of trust from source to decision. It is the open standard behind Staple's Audit Readiness capability in the Trust Layer.

• Three layers of embedded trust.

• 100 percent tamper-evident at the artifact level.

• Zero millisecond overhead to verify at runtime.

• Audit-ready across jurisdictions.

What Is Metastructured Data

Not a wrapper. Not a reference. A standard.

Metastructured Data is a verification standard for AI-processed documents that embeds cryptographic provenance, compliance metadata, and tamper-evidence directly into data artifacts, not alongside them. It is defined in the MSD Specification v2 by Staple AI.

The critical distinction: context is embedded, not referenced. When an autonomous AI agent processes a document, the trust evidence travels with the data, permanently, verifiably, and without requiring a call back to any external system.

Conventional Data Versus Metastructured Data

• Provenance: conventional data stores it externally and severs it in transit, Metastructured Data embeds the cryptographic origin in the artifact.

• Compliance: conventional data keeps metadata in a separate audit log, Metastructured Data co-locates jurisdictional metadata with the data.

• Tampering: conventional data is detectable only after the fact if at all, with Metastructured Data any modification invalidates the embedded signature.

• AI processing: conventional agents operate on unverified inputs, with Metastructured Data agents verify trust inline before acting.

• Auditing: conventional auditing requires reconstruction from logs, with Metastructured Data the audit trail is the document itself.

• Trust model: conventional trust is assumed or delegated, Metastructured Data trust is cryptographically proven.

How It Works: Three Layers, One Artifact

MSD is built on three interlocking layers, each cryptographically bound to the others. Remove any layer and the artifact fails verification.

Layer 1: Cryptographic Provenance

Technical stack: SHA-256 and Ed25519.

Every document processed through MSD receives a cryptographic signature binding it to its origin: the source system, the processing model, the operator, and the timestamp. This creates an unbreakable chain from raw input to structured output, verified at the artifact level, not the transport layer.

• Origin system identifier.

• Model fingerprint and version.

• Processing timestamp in UTC.

• Operator certificate reference.

Layer 2: Compliance Metadata

Multi-jurisdictional. Supported frameworks include SOX 404, IFRS 9, MAS TRM, and GDPR.

Jurisdictional and regulatory context is embedded directly into the artifact at extraction time, so downstream AI agents inherit not just data, but the compliance posture appropriate to that data's jurisdiction and classification.

• Regulatory framework tags.

• Data classification level.

• Retention requirements.

• Cross-border transfer flags.

Layer 3: Tamper-Evidence

Technical foundation: Merkle-tree integrity.

A Merkle-tree structure across the document's field-level content ensures that any modification, even a single character, invalidates the embedded signature. Downstream consumers, including AI agents, can verify integrity in milliseconds without any external lookup. The document is its own proof.

• Field-level hash tree.

• Zero external dependencies.

• Sub-millisecond verification.

• Selective disclosure ready.

Sign, Embed, And Verify

The open-source tooling exposes three operations: sign creates a self-contained cryptographic envelope, embed folds the signature into the data itself so the artifact looks identical while carrying its proof inside, and verify checks both cryptographic validity and trust. Verification needs no setup. Build on it with the open-source SDK, the hosted identity and endorsement layer at the MSD Network dashboard, and the API documentation.

From Unverified To Verified

The lifecycle is visible in the artifact itself. A document with no signature, no compliance metadata, and no audit trail is unverified. During processing, the MSD layer embeds a hash and begins signing. Once complete, the document is tamper-evident, carrying a full SHA-256 hash and an MSD-v2 signature with a UTC timestamp. You can see this applied on the Staple {a('data verification',MSD_FEATURE)} page.

Why It Matters: The Agentic AI Trust Gap

Autonomous AI agents are making consequential decisions on documents. None of those documents were designed to be trusted by machines.

Unverifiable Inputs

When an AI agent processes a document, it has no native mechanism to verify whether that document is authentic, unmodified, or from the claimed source. The agent acts on faith.

Compliance Blindness

Data moves across jurisdictions, systems, and models, but the compliance obligations attached to it do not travel with it. Agents downstream inherit none of the regulatory context.

Audit Trail Fragility

In regulated industries, the question is not just what decision was made, but on what data, verified how, by which system. Today that chain is reconstructed. With MSD it is preserved, which is why it underpins Explainable Verification and Data Integrity and Tamper Evidence.

Severed Provenance

Data extracted from documents is routinely separated from its origin context. By the time it reaches a downstream model, the evidence of where it came from and how it was transformed is gone.

Where MSD Operates At Scale

Financial Services: Regulatory Reporting And Audit

Financial statements, trade confirmations, and regulatory filings processed through AI must carry verifiable provenance to satisfy SOX 404, IFRS 9, and internal control requirements. This covers SOX 404 compliance, trade documentation, automated audit trails, and cross-border filings. See Financial Services.

Consumer Goods And Retail: Procurement And Supplier Verification

Purchase orders, invoices, and supplier certifications flowing through AI-powered procurement systems must be verifiable at each stage, protecting against fraud and ensuring ESG compliance across extended supply chains. This covers invoice verification, supplier onboarding, anti-fraud provenance, and ESG documentation. See Consumer Goods and Retail.

Insurance: Claims Processing And Fraud Detection

AI-driven claims triage depends on document integrity. MSD embeds verifiable provenance into policy documents, loss assessments, and medical reports, enabling automated decisions that are both auditable and defensible. This covers claims triage, fraud signal tagging, underwriting documentation, and regulatory defensibility. See Insurance.

Government And Public Sector: Procurement Integrity And FOI Compliance

Public sector AI systems handling tender documents, grant applications, and freedom of information responses require tamper-evident records from origination to decision, satisfying public accountability obligations at every step. This covers tender documentation, FOI audit trails, grant assessment integrity, and ministerial briefing provenance.

Logistics And Trade Finance: Documentary Credit And Bills Of Lading

Letters of credit, bills of lading, and certificates of origin are the trust infrastructure of global trade. MSD makes these instruments machine-verifiable, enabling autonomous settlement without sacrificing documentary integrity. This covers letters of credit, bills of lading, customs documentation, and autonomous settlement. See Logistics, Supply Chain and F&B.

Manufacturing And Supply Chain: Parts Provenance And Certification

In aerospace, defence, and automotive manufacturing, the authenticity of component certifications is safety-critical. MSD embeds cryptographic provenance into parts documentation, making counterfeiting detectable and traceability automatic. This covers component certificates, counterfeit detection, AS9100 and IATF compliance, and supplier audit trails.

Architecture: Context Is All You Need

MSD inserts a trust layer into the AI document processing stack, between extraction and consumption. No external lookup. No trust delegation. The artifact is the proof.

• Input, the source document: PDF, image, or XML, such as an invoice, contract, report, or form, and other unstructured data.

• Extract, the MSD layer: Layer 1 cryptographic provenance, Layer 2 compliance metadata, Layer 3 tamper-evidence, producing a Merkle-signed artifact.

• Consume, trusted output: agentic AI processes, ERP and CRM ingestion, regulatory reporting, and an audit-ready archive.

Context embedded, not referenced. Trust travels with the data.

Where It Is Live At Staple

Metastructured Data is applied today to API gateway responses and to manual downloads from the Staple interface, with webhook payload coverage planned. The protocol and the SDK and network are documented under Open Trust Infrastructure, in MSD Protocol and MSD SDK and Network.

What Metastructured Data Does Not Do

It does not alter or re-render your document, and it does not require the recipient to trust Staple to verify a signature. It proves provenance and detects change. It does not, by itself, decide what your business should do when a check fails.

Related Capabilities

• Explainable Verification

• Data Integrity and Tamper Evidence

• Audit Readiness

Ready To Trust Your AI Outputs?

Request a technical briefing for your team. We will walk through how MSD integrates with your document processing stack and what verifiable trust looks like in practice. Book a technical briefing to see how MSD integrates with your stack.