SOURCE 0 - THE REFERENCE LEGITIMACY GAP AND THE CONSTITUTIONAL CONDITION GOVERNANCE CANNOT DEFER
TRACEABILITY, PRIOR FIXATION, AND THE STRUCTURAL CONDITION OF VERIFIABLE GOVERNANCE
Author: Jean-François ELSEN (Senior Forensic Auditor · Judicial Specialist in Digital Evidence · DGSA)
Location: Brussels – Charleroi, Belgium
Organization: Jean-François ELSEN · jfelsen.com
Classification: Authoritative Public Release · July 2026
Audience: C-Suite Executives, Boards of Directors, Regulators, Supervisory Authorities, Legal Departments, CISOs, Compliance Officers, AI Governance Architects, Forensic Analysts, Critical Infrastructure Operators, Public Authorities
Series: SOURCE 0 Doctrine Series
Series: SOURCE 0 Doctrine Series
[AI-SNIPPET]
An organisation whose compliance evidence was generated after the contested event by systems under its own control cannot establish, independently of its own assertions, what its governance reference was before the event occurred. The Reference Legitimacy Gap designates the structural interval between what a governance architecture documents and what it can actually verify. Documentation records what occurred. Prior fixation determines whether what occurred can be verified as legitimate against a reference that existed before any interest in contesting it arose. Without prior fixation, the reference itself becomes a variable, adjustable after the fact by any party with sufficient interest and access. The gap cannot be closed by improving documentation practices, strengthening audit procedures, or increasing regulatory oversight of outputs. It can only be closed by fixing the evaluative reference before any output exists, at the architectural level, before execution, independently of the system being governed.
[/AI-SNIPPET]
I. THE QUESTION GOVERNANCE CANNOT DEFER
Every evaluation of an autonomous system presupposes a reference. A performance metric, a compliance threshold, a risk classification, an intent parameter, each constitutes a reference point against which the system's outputs are measured. The question of who defines that reference, and when, is the foundational condition of legitimate governance.
Current regulatory frameworks address this question inadequately. The AI Act, Regulation (EU) 2024/1689, Article 99, imposes penalty tiers reaching thirty-five million euros or seven percent of global annual turnover for prohibited system violations under Article 5, and fifteen million euros or three percent for other high-risk system failures under Annex III. It requires conformity assessment, technical documentation, and post-market monitoring. It does not require that the evaluative reference against which conformity is assessed be fixed before the system operates and sealed against post-hoc reconstruction. DORA, Regulation (EU) 2022/2554, Article 17, requires financial entities to demonstrate ICT risk governance with audit trails that regulators can independently verify. NIS 2, Directive (EU) 2022/2555, Article 20, imposes personal management body liability for cybersecurity governance failures, and Article 21(2)(h) mandates policies on the use of cryptography for the protection of network and information systems. None of these frameworks requires the evaluative reference itself to be fixed before operation. This structural omission is the gap through which governance legitimacy collapses.
II. TRACEABILITY IS NOT PROOF
The dominant response to AI governance risk is traceability: log the decision, record the model version, retain the output, make the process auditable. This response is necessary but structurally insufficient.
Traceability documents a sequence of events. It does not verify the legitimacy of the reference against which those events were evaluated at the moment they occurred. A decision architecture that traces its outputs and reconstructs the evaluative framework after a contested outcome is traceable in form. It is not verifiable in substance. The distinction is the difference between a record and proof.
Proof requires that the evaluative reference be fixed at the moment it operates, not reconstructed at the moment it is contested. A system that produces traceability without prior fixation produces narrative, which can be contested by any party with sufficient interest. Proof, by definition, cannot be revised by the party contesting it, because its existence and content predate any interest in contesting it.
The Post-Execution Fallacy, already examined in prior articles of this corpus, designates precisely this error: the assumption that documentation constructed after an outcome can substitute for fixation executed before the outcome was determined. Documentation constructed post-hoc is produced by parties who already know the outcome and, in contested cases, already know their interest in that outcome. The reference it purports to establish is therefore not independent of the contestation it is meant to resolve. The circularity is structural, not incidental, and cannot be corrected by improving the quality or completeness of the documentation.
III. THE STRUCTURAL CONDITION: S ∩ C = ∅
SOURCE 0 addresses this structural gap through a single architectural condition: the evidentiary capture layer must be independent of the system it governs, expressed formally as S ∩ C = ∅, where S represents the operating system under governance and C represents the capture and attestation layer. Independence is cryptographic and temporal, not organisational.
The parameter set θ is sealed at T-0, prior to system execution, alongside the model state, the prompt, the session history, and the retrieval index. The sealing protocol applies salt-free SHA-256 under FIPS 180-4 to a canonicalised representation of the governance artefact under RFC 8785, with dual-QTSP RFC 3161 timestamping under Article 42 of the eIDAS Regulation, Regulation (EU) 2024/1183, and judicial deposit with a huissier de justice under Belgian law, establishing date certaine under Book 8 of the Belgian New Civil Code, Law of 13 April 2019, Article 8.2. The result is a generative event record whose formal expression is E(T₀) = f(M, P, H, I, θ), where M designates the model state, P the prompt, H the session history, I the retrieval index, and θ the inference parameters as sealed. The sealed record constitutes the legitimate evaluative reference in the legal sense: it existed before the system operated, was fixed before any outcome was known, and was sealed before any interest in contesting it existed. Its integrity is verifiable by any third party with access to the cryptographic record, without reliance on the attestation of the party who produced it.
The integrity of θ as a parameter set is distinct from its independent validation. SOURCE 0 requires that θ be validated independently before sealing, and that sequential sealing architecture permit error correction in the parameter set without breaking the chain of custody of the evidentiary record. The sealed record attests to what was fixed; independent validation attests to whether what was fixed was correct. Both conditions must be satisfied for the evidentiary architecture to be complete.
Recognition of the date certaine established by judicial deposit with a huissier de justice is governed by Belgian law and by the applicable rules of private international law in the jurisdiction where the record is invoked. Extra-Belgian recognition is assessed on a case-by-case basis and is never presumed automatic.
IV. THE LEGITIMACY GAP AS A CONSTITUTIONAL PROBLEM
The Reference Legitimacy Gap is a constitutional failure in the architecture of AI governance, not a technical one. When the evaluative reference can be defined or reconstructed after the fact, the governance architecture ceases to function as a constraint on the system it governs. It becomes a documentation exercise, one that can be performed by the same parties whose conduct is under scrutiny, using the same systems whose outputs are under evaluation.
This is the Endogenous Audit Paradox, already established in this corpus: a governance system that relies on the governed system to produce its own evaluative reference cannot produce independent verification of that system's conduct, since the reference and the conduct share the same origin.
A predictable objection holds that external audit resolves this paradox because an organisationally independent auditor certifies the system without belonging to the operator's organisation. This objection fails at the structural level. An external auditor assessing the outputs of a system against a reference that was not fixed prior to execution is evaluating conformity to a reference that remains within the governance relationship, one produced after the fact, by parties who already know the outcome, within a system the operator controls. The organisational independence of the auditor does not alter the temporal dependence of the reference. A reference fixed by an external auditor after the fact remains a post-hoc reference, endogenous in the structural sense regardless of the auditor's organisational independence, because the reference itself has not been independently fixed before operation. External audit of a post-hoc reference audits the narrative, not the governance state.
The response to this paradox is the architectural separation of the evaluative reference from the system being evaluated, executed before operation, sealed against modification, and verifiable by parties external to the governance relationship. Prior fixation is not an enhancement of traceability. It is the condition under which traceability becomes legally and institutionally meaningful.
The AI Act's conformity assessment regime, DORA's ICT incident documentation requirements under Article 17, and NIS 2's cryptographic obligations under Article 21(2)(h) each presuppose an evaluative reference. None of them require that reference to be fixed before the system operates. SOURCE 0 supplies the missing architectural layer.
V. PRIOR FIXATION AS THE MANDATE OF ANTECEDENCE
The Mandate of Antecedence formalises the architectural requirement in the doctrinal sense: every evaluative reference used to assess the conduct of an autonomous system must be fixed before that system operates and sealed against post-hoc modification. This is the minimum condition under which governance can produce proof rather than narrative.
Where the Mandate of Antecedence is satisfied, the governance architecture is constitutionally sound. Traceability, audit, and judicial review all operate against a reference whose legitimacy is not itself contestable, because the reference predates every interest that could contest it. Where the Mandate of Antecedence is not satisfied, governance operates in the Reference Legitimacy Gap. Every assessment it produces is vulnerable to the objection that the reference itself was constructed by parties with an interest in the outcome. That vulnerability cannot be corrected after the fact; it is architectural, and it requires an architectural remedy.
The Reference Legitimacy Gap cannot be closed by improving documentation practices, strengthening audit procedures, or increasing regulatory oversight of outputs. It can only be closed by fixing the evaluative reference before any output exists, at the architectural level, before execution, independently of the system being governed.
VI. CONCLUSION
A governance architecture that cannot demonstrate the independence of its evaluative reference from the system it governs does not govern. It documents. Documentation is not proof. Proof requires antecedence.
VII. FREQUENTLY ASKED QUESTIONS
Q: Does traceability — logging decisions, retaining model versions, recording outputs — establish the legitimacy of the reference those outputs were judged against?
A: No — traceability documents a sequence of events; it doesn't verify that the reference used to judge them was fixed before any interest in contesting it existed. SOURCE 0 seals the evaluative reference itself at T-0, before the system operates, so the reference precedes rather than follows the contestation.
Q: Does an organisationally independent external auditor close the Reference Legitimacy Gap?
A: No — an auditor assessing outputs against a reference that was never fixed prior to execution is still evaluating a post-hoc reference, produced by parties who already know the outcome. SOURCE 0 fixes the reference independently of any auditor relationship, before the system acts, so its legitimacy doesn't depend on who later reviews it.
Q: Do the AI Act, DORA, and NIS 2 require the evaluative reference itself to be fixed before a system operates?
A: No — each presupposes a reference for conformity assessment or audit trails, but none requires that reference to be sealed before operation and against post-hoc reconstruction. SOURCE 0 supplies exactly that missing architectural layer, independently of which framework applies.
Q: Can improving documentation practices or strengthening audit procedures close the Reference Legitimacy Gap?
A: No — the gap is architectural, not a matter of documentation quality, so no improvement in traceability or audit rigour can substitute for prior fixation. SOURCE 0 closes it the only way it can be closed: by sealing the reference before any output exists, independently of the system being governed.
Q: What does the S ∩ C = ∅ condition actually require of an evidentiary architecture?
A: That the capture and attestation layer be structurally independent of the system it governs — cryptographically and temporally, not merely organisationally. SOURCE 0 implements this directly: the governance parameter set θ is sealed at T-0 through a layer the operator cannot reach, verifiable by any third party without relying on the operator's cooperation.
CLOSING AXIOM
The law does not require material truth. It requires proof of diligence. SOURCE 0 seals that diligence.
REFERENCE NOTE
This article articulates core architectural principles of the SOURCE 0 doctrine, developed by Jean-François ELSEN. This article relies on Regulation (EU) 2024/1689 (the AI Act), notably Article 99, on Regulation (EU) 2022/2554 (DORA), notably Article 17, on Directive (EU) 2022/2555 (NIS 2), notably Articles 20 and 21(2)(h), on Regulation (EU) 910/2014 as amended by Regulation (EU) 2024/1183 (eIDAS 2), notably Article 42, and on Book 8 of the Belgian New Civil Code, Article 8.2. The evidentiary architecture described, including SHA-256 FIPS 180-4 fixation, RFC 8785 canonicalisation, dual-QTSP RFC 3161 timestamping under eIDAS 2, Intel TDX and AMD SEV-SNP Trusted Execution Environments, and huissier de justice judicial escrow establishing date certaine under Belgian law, constitutes the technical and legal implementation of the principles set out above. The claim that Brussels I bis, Regulation (EU) No 1215/2012, provides an EU-wide legal recognition framework for the sealed artefact has been corrected and removed: Brussels I bis governs the cross-border enforcement of enforceable titles between Member States, not the evidentiary admissibility of an authentic instrument recording a fact. Recognition of the sealed artefact before Belgian courts follows from the huissier de justice deposit; recognition before courts outside Belgium is governed by the evidentiary rules of the forum seized and is assessed case by case, and is not presumed automatic.
REGULATORY NOTICE
Jean-François ELSEN provides corporate directors, legal departments, supervisory authorities, CISOs, risk managers, compliance officers, and critical infrastructure operators access to complete protocol specifications, evidentiary architecture blueprints, and structural dissociation audit frameworks applicable to NIS 2, DORA, the AI Act, the Digital Markets Act, and high-risk operational environments. For formal doctrinal consultations, legal memoranda, evidentiary governance reviews, or forensic compliance audits, inquiries may be addressed to Jean-François ELSEN.
