Technical Whitepaper24 pagesPublished December 2025

Post-Quantum Cryptography and Sealed Evidence for AI Sovereign Systems

A design-pattern overview of how post-quantum cryptography and SHA-256 Merkle-sealed evidence records could support provenance of AI models and training data. It describes design options, not a shipped product feature.

Request a Demo

Free Preview

Executive Brief

AI Sovereignty Requires Verifiable Records

As AI systems become critical infrastructure, organizations need verifiable records of model provenance and training data integrity. Without them, AI systems remain exposed to supply chain attacks, data poisoning, and model tampering that nobody can detect after the fact.

This whitepaper describes a design pattern for AI sovereignty that combines post-quantum cryptography with Merkle-sealed evidence records, so AI systems can be audited in the quantum computing era. It describes design options, not a shipped product feature.

ML-DSA

Design option

via liboqs (not CAVP-validated)

Tamper-evident

Audit Trail

Records sealed under a Merkle root

Design pattern

Approach

Not a shipped feature

EU AI Act

Discussed

Not a certification

What's Inside the Full Whitepaper

AI Supply Chain Security

Content-hash records of model origins and training pipeline artefacts, as a design pattern.

Training Data Provenance

Tamper-evident records of data sources, preprocessing, and augmentation steps, sealed under a Merkle root.

Model Signing Options

Post-quantum signing options (ML-DSA-65 via liboqs, not CAVP-validated) for model artefacts, as a design pattern.

Inference Records

A design pattern for recording which model version served an inference, sealed under a SHA-256 content hash.

Compliance Integration

Discussion of the EU AI Act and NIST AI RMF and how they relate to provenance records.

Implementation Architecture

A reference architecture sketch for AI provenance records, described as a design pattern.