Introduction — Why this Atlas exists

Introduction

Three force-multipliers landed on the trust-services regime in roughly the same eighteen months: the EU AI Act, the NIST post-quantum standards, and AI agents that act on behalf of people. None of the three came with a unified practitioner’s guide for the engineers who already deploy eIDAS-grade infrastructure. This atlas is one such guide, in three languages, with audio.

0.1 Why this book exists

PKI literature in English is mature but fragmented. PKI literature in Estonian and Russian is sparse, scattered, and frequently outdated. At the same time, two new force-multipliers have arrived in the trust domain in 2024–2026:

• Post-quantum cryptography migration. NIST FIPS 203 (ML-KEM), FIPS 204 (ML-DSA), and FIPS 205 (SLH-DSA) became standards in 2024, kicking off the largest cryptographic migration in the history of the Internet. The migration will occupy PKI engineers for the rest of the decade.
• AI agents under regulation. The EU AI Act (Regulation 2024/1689) entered into force in 2024 and applies in stages between 2025 and 2027. Articles 9–15 (the high-risk regime) make demands of integrity, transparency, record-keeping, and human oversight that the existing eIDAS / ETSI EN 319 trust services regime is, in significant part, already prepared to meet — if you know how to map the obligations onto the primitives.

This atlas is one practitioner’s attempt to close those gaps in three languages, simultaneously. Where prior PKI books focused on a single audience (compliance officers, web-PKI engineers, smart- card developers), this one explicitly targets the intersection of PKI, post-quantum migration, and AI-agent governance. The bias is toward operationalisable content — what a competent engineer can deploy on Monday morning, not what a working group has been arguing about for fifteen years.

The book is not a substitute for the standards. It is a reading guide for them. Where ETSI EN 319 102-1 specifies what a signature creation procedure must do, we tell you which sections to read, which footnotes are load-bearing, and which options are operationally defensible in 2026.

0.2 Who this book is for

The atlas is written for four audiences, in roughly decreasing proportion of expected readership:

1. Engineers who deploy or maintain trust infrastructure — QTSP operators, internal PKI teams, embedded device security developers, anybody who has ever woken up at 03:00 to a “certificate validation failure” alert. Chapters 1, 6, and the partner-integration appendices are for you.
2. Architects designing systems that consume trust infrastructure — application architects, identity-provider designers, AI-platform architects. Chapters 2, 4, and 7 are the spine of your reading.
3. Auditors and project managers translating between compliance language and engineering reality — particularly under the EU AI Act high-risk regime. Chapters 5 and 7 carry most of the load; the glossary in the back is your friend.
4. Curious readers with a security background who want to understand what the post-quantum / AI-Act conversation is actually about. Chapter 0 (this one), 3, and 8 are written with you in mind.

What the four audiences share is a working vocabulary: certificates, hashes, signatures, TLS, OAuth, JWT. We explain anything beyond that working vocabulary as it appears.

0.3 How to read this book

The chapters are ordered for sequential reading, but each is designed to stand alone for reference. A reader who already knows PKI may skim chapter 1, dwell on chapter 3, skip directly to chapter 5, and use chapter 6 as a cookbook. A reader new to trust services should read in order; chapter 1 is non-negotiable for anything that follows.

Chapter	What it gives you
0 — Introduction	The premise. (You are here.)
1 — PKI Fundamentals	The substrate everything else assumes.
2 — eIDAS and ETSI EN 319	The European regulatory frame.
3 — Post-Quantum Migration	The next decade’s transition.
4 — AI Agents in the Trust Stack	The three-layer model: identity / provenance / governance.
5 — EU AI Act for PKI Engineers	Articles 10/12/13 read in PKI-native terms.
6 — Operational Playbooks	TSA, OCSP, audit ledger, EATF flow as runnable recipes.
7 — Sector Verticals	The substrate in production, across five sectors.
8 — Open Problems	Where the PKI lens stops being sufficient.

Cross-references are explicit: where chapter 5 builds on chapter 4, the link is in the prose, not implicit. The book is published as web text and as audio; the audio is read in the author’s own voice, cloned through MiMo-V2.5-TTS-VoiceClone. Listen if you prefer it; read if you don’t.

0.4 What this book is NOT

It is important to be specific about what this atlas does not attempt, because the trust-services / AI-Act space attracts overclaiming.

• A legal text. The EU AI Act mapping in chapter 5 is an engineering reading, validated against the official Regulation but not a substitute for qualified counsel. Where we say “this obligation maps onto this primitive”, we mean as a matter of engineering, not as a matter of European law. Lawyers sometimes disagree with engineers; in those disagreements, lawyers win.
• A standards document. We cite ETSI EN 319, NIST FIPS, IETF RFCs, and the AI Act itself. We do not redefine them. We occasionally interpret them, which we mark as interpretation.
• A vendor pitch. We use Aletheia / EATF as a reference implementation in the worked examples because the author wrote it. The atlas would still hold if a different signing engine were substituted, and we deliberately structure the sections so that substitution is straightforward.
• A complete catalogue of failures. The “common mistakes” sections in chapters 1 and 6 enumerate the failures the authors have personally seen in production. They do not enumerate every possible failure; the reader’s deployment will have its own.

If you came looking for the Court of Justice of the EU’s reading of Article 14, or for a NIST Special Publication on PQC migration, this is the wrong book. We point you to the right ones in the references.

0.5 Companion materials

The atlas is one of three artefacts that share a research corpus:

• The atlas itself (this book) — long-form, practitioner- oriented, trilingual. Audio version available once recordings are calibrated.
• The preprint — “Cryptographic Provenance for AI Outputs: a PKI-Native Reading of EU AI Act Articles 10/12/13 through eIDAS and ETSI EN 319.” The academic version of chapter 5, with formal apparatus, threat model, and CDDL schema for the .aep Evidence Package. Hosted at IACR ePrint and on the paper-pki-ai-act repository.
• The reference implementation — Aletheia / EATF. Open-source, runnable, with offline .aep verification.

Every artefact in this atlas — every chapter, every audio file, every diagram — is signed by Aletheia. A QR code at the end of each chapter links to atlas.h2oatlas.ee/verify/<chapter-id>, where you can independently verify the chapter you just read came from the author and has not been tampered with. We think this is worth doing for two reasons. First, it lets us hold ourselves to the standard the book argues for. Second, it provides a small, friendly worked example of the substrate-meets-vertical pattern in chapter 7.

0.6 Voice and conventions

The atlas is written in deliberate, declarative prose. We use:

• Terminology consistent with the glossary in three languages — RIA Estonian, ГОСТ Russian, ETSI / IETF English. Where a term has no good translation, we say so and use the original.
• Code, hex, and ASN.1 fragments where they aid understanding; we never reproduce wire-format dumps for their own sake.
• Diagrams for any concept that has more than three moving parts. The diagrams in this book were rendered through MiMo-V2-Omni from plain-text schematics; the schematics are in the diagrams/ directory of the source repository, so a reader who disagrees with a diagram can fork and re-render.
• Footnotes sparingly. If a footnote would clarify, the prose probably needed work.
• Estonian asides when the local production picture illuminates a global concept. The author lives in Estonia and has run PKI infrastructure here; the local examples are first- hand, not stereotypical.

Acknowledgements are at the end of each major contribution (chapter 5, chapter 7) where the contribution is from outside the author. The book itself is the author’s, with explicit co-authorship from Claude Opus 4.7 (drafting, structuring) and MiMo-V2.5-Pro (Estonian and Russian translation; activated under the Xiaomi MiMo 100T grant of 2026-05-05).

0.7 A reading order for the impatient

If you have an hour and want the core argument:

1. Skim §1.2 (X.509 anatomy) and §1.4 (revocation).
2. Read §4 (the three-layer model — identity, provenance, governance) end-to-end.
3. Read §5 (the AI Act mapping).
4. Read §8 (open problems).

If you have a weekend, read in order. If you have a project, the glossary and chapter 6 will save you weeks.

Welcome.