What Pharma, ESG, and Media Teach Us About Enterprise Provenance — And Where the Gaps Still Are

The Pattern That Keeps Appearing

Across industries that look nothing alike — pharmaceutical supply chains, corporate sustainability reporting, and digital journalism — organizations have independently reached the same conclusion: tracking that something happened is not the same as verifying that it happened correctly, by the right party, at the right time.

The pharmaceutical industry learned this during the DSCSA compliance rollout. ESG teams learned it when "sustainable sourcing" claims started attracting regulatory scrutiny. Media organizations learned it when AI-generated images began circulating as photojournalism.

Each industry built different infrastructure to solve the problem. Each hit different limits. The gaps they expose are instructive for any organization building content provenance today.

Pharmaceuticals: MediLedger and the DSCSA Compliance Problem

The Drug Supply Chain Security Act (DSCSA) requires pharmaceutical manufacturers, distributors, and dispensers to track and verify prescription drugs at the unit level as they move through the supply chain. The compliance deadline has driven significant investment in serialization infrastructure — barcodes, EDI systems, and, increasingly, blockchain-based verification networks.

MediLedger (now Chronicled) built one of the most mature pharmaceutical blockchain networks in production. Their system allows trading partners to verify product identifiers against a shared ledger without exposing proprietary data to competitors. The key insight driving their architecture: a barcode scan proves a product was present at a location. A cryptographic verification proves the trading partner presenting that product is who they claim to be and has an authorized chain of custody.

This distinction — between tracking presence and verifying identity — is the same distinction that separates a hash registry from a signed provenance record.

Where it hits limits

MediLedger's network is permissioned and pharmaceutical-specific. Onboarding requires bilateral trading partner agreements. Verification is network-internal — a pharmacist cannot independently query the ledger the way a regulator can query a public endpoint. The compliance reports that DSCSA requires are still largely assembled manually from system exports.

The core lesson for digital content pipelines: cryptographic verification at every handoff is the right architecture. The operational overhead of a permissioned network with manual compliance reporting is the avoidable part.

ESG Reporting: Unilever, GreenToken, and the Self-Reported Data Problem

Unilever's pilot with GreenToken by SAP (2022) was one of the most visible enterprise attempts to apply blockchain to ESG supply chain data. The goal was to track sustainable palm oil sourcing — a commodity with significant deforestation risk — from certified mills through to finished products. The blockchain layer was meant to give stakeholders verifiable proof that sustainability claims corresponded to real sourcing decisions.

The pilot demonstrated something important: when environmental metrics are anchored to a blockchain at the point of measurement — mill certification, harvest batch, logistics handoff — the data is meaningfully more trustworthy than spreadsheet-based self-reporting submitted weeks later. The immutability of the ledger prevents retroactive adjustment of inconvenient numbers.

Where it hits limits

GreenToken and similar ESG blockchain systems solve the data integrity problem downstream of the entry point. If the data entered at the source is wrong — a mill over-reports certified volume, a supplier misattributes a batch — the blockchain faithfully records the incorrect data and makes it tamper-evident. Garbage in, immutable garbage out.

The deeper problem is that ESG data entry is typically a manual, periodic process conducted by people with incentives to report favorably. Signing at the source only helps when "the source" is an automated measurement system rather than a human filling in a form.

For digital content pipelines, this lesson translates directly: sign at the moment of generation, not at the moment of submission. An AI-generated image signed at the point the model produces it is a stronger provenance claim than the same image signed three days later when someone uploads it to a DAM system. The gap between creation and signing is where the claim weakens.

Media and Journalism: The NYT News Provenance Project and Reuters/Starling Lab

The New York Times News Provenance Project and the Reuters/Starling Lab authenticity proof of concept are among the most carefully designed content provenance implementations in production. Both are built on C2PA — the Coalition for Content Provenance and Authenticity standard — which embeds a cryptographically signed manifest directly into media files. The manifest records the capture device, the capture time, any edits made, and whether AI tools were involved.

C2PA is technically sound. When a photojournalist captures an image on a C2PA-enabled camera, the manifest is written at capture time and signed by the device. Any subsequent modification — cropping, color correction, AI enhancement — is recorded in the manifest's edit history. A reader or editor can verify the chain of custody from sensor to publication.

The Starling Lab project went further, combining C2PA with decentralized storage (IPFS, Filecoin) and blockchain timestamping to create archival-grade provenance for conflict journalism. The goal was records that would be verifiable in legal proceedings and historical archives, independent of any single organization's continued existence.

Where it hits limits

C2PA solves content authenticity — proving a specific image is what it claims to be. It does not solve operational provenance for enterprise content pipelines.

• •C2PA is embedded in the file. If the file is stripped of metadata during compression, CDN delivery, or social platform upload — which happens routinely — the manifest is gone.
• •C2PA handles image and video well. Text documents, audio, and multi-modal AI outputs have inconsistent support.
• •C2PA does not provide a unified audit trail across multiple tools and pipelines. A brand producing AI-generated content across Firefly, Midjourney, and an internal model gets three separate provenance systems with no common query interface.
• •C2PA does not address the business records, licensing agreements, and contractual obligations tied to an asset — only the asset's content history.

The Reuters/Starling approach of anchoring records externally — decoupled from the file itself — is architecturally closer to what enterprise pipelines need. The record lives in an independent ledger. The file carries a reference. Verification queries the ledger, not the file.

The Common Architecture That Works

Across all three industries, the provenance implementations that hold up under regulatory scrutiny share the same architectural properties:

• •Sign at the source. MediLedger verifies at each supply chain handoff, not in a nightly batch. C2PA captures at the camera shutter, not at the editor's desk. The earlier in the process the signature is applied, the stronger the provenance claim.
• •Write to an immutable ledger. Every system that has survived audit scrutiny uses an append-only record store. The inability to modify past records is the property regulators are actually checking for.
• •Expose a public verification endpoint. MediLedger's network is permissioned, which limits who can verify. C2PA embedded in a file depends on the file surviving intact. The implementations that work best for external audit make verification available to anyone with the record identifier — no prior relationship, no authentication required.
• •Separate the record from the asset. File-embedded provenance is fragile. An external ledger record that references the asset by hash survives any format conversion, compression, or platform upload the asset goes through.

How This Maps to the Ledgible Architecture

The pharmaceutical, ESG, and media industries each spent years and significant engineering resources discovering that signing at the source, writing to an immutable ledger, and exposing public verification are the non-negotiable properties. Ledgible is built with those properties as defaults, not as features to be added later.

For an enterprise content pipeline — regardless of industry — the practical implication is the same: one API call at generation time produces a cryptographically signed, publicly verifiable provenance record that satisfies regulatory requirements across jurisdictions, survives any downstream processing the asset goes through, and can be queried by any external party without access to your internal systems.

What Each Industry's Lesson Means for Your Pipeline

If you are in a regulated industry

The DSCSA experience demonstrates that regulators will eventually require cryptographic verification, not just tracking logs. Building provenance infrastructure before the compliance deadline — rather than scrambling to retrofit it — is the difference between a one-click compliance report and a week-long audit. The organizations that built early had working systems when enforcement began. The ones that waited built under pressure.

If you produce ESG or sustainability data

Self-reported metrics are under increasing scrutiny from regulators (SEC climate disclosure rules, EU CSRD), investors, and the public. The GreenToken pilot showed that blockchain anchoring of metrics at the point of measurement is technically feasible and meaningfully more trustworthy than spreadsheet submission. For organizations producing AI-assisted sustainability analysis or generating reports from automated data pipelines, signing that data at generation time is the same principle applied to a different asset type.

If you produce AI-generated content at scale

The NYT and Reuters work shows that the media industry has already accepted that content provenance infrastructure is necessary infrastructure, not a nice-to-have. C2PA solves the file-level authenticity problem for image and video. It does not solve the enterprise pipeline problem. The gap between what C2PA covers and what EU AI Act Article 50 requires — machine-readable disclosure across all asset types, across all tools, with a unified audit trail — is the gap Ledgible is built to close.