Version 2 – 27 April 2026 – 16:49 UTC – Extension of the initial formulation recorded on 27 April 2026
Previous version: EVIDE Boundary Model — First recorded formulation (27 April 2026)
Canonical Formulation
Layer 1 determines validity of attribution.
Layer 2 determines strength of attribution.
EVIDE Boundary Model — Two-Layer Architecture
The EVIDE Boundary Model defines the transition from system-internal evidence to externally attributable responsibility. This transition is not triggered by data completeness alone, but by structured responsibility conditions organized across two distinct levels.
Level 1 — Handoff Validity
Level 1 determines whether an object can cross the Layer 2 to Layer 3 boundary.
Conditions:
- Behavioral completeness
- Responsibility closure
- Authority coherence
If these conditions are not satisfied simultaneously, the object is rejected at the boundary. Level 1 answers the question: can responsibility be attributed?
Level 2 — Evidentiary Strength Layer
Level 2 determines how strong the object remains once outside the originating system. This is not a gate, but a measure of evidentiary strength.
Conditions:
- Interpretability — classification context stability
- Reproducibility — trace continuity
- Governance transparency — threshold reference clarity
Level 2 answers the question: is that responsibility defensible outside the system?
Key Distinction
An object may satisfy Level 1 and still present degraded evidentiary strength at Level 2. In legal and audit contexts, a passed-but-weak object is often more problematic than a rejected one, because it creates ambiguity rather than clarity.
Critical Insight
The EVIDE Boundary Model shifts AI governance from demonstrating what happened to establishing who is responsible and how defensible that responsibility remains once it leaves the originating system boundary.
Extended Validation — Graded Evidentiary States
Further validation of the EVIDE Boundary Model has introduced a third operational state beyond binary acceptance and rejection.
The model now distinguishes between three outcomes:
- Rejected at Level 1 — attribution is not valid, object does not cross the boundary
- Passed with full evidentiary strength — attribution is valid and fully reproducible
- Passed with degraded evidentiary strength — attribution is valid but the responsibility chain is not fully reproducible outside the system
This extension confirms that Level 2 is not a secondary gate, but a structured measure of evidentiary strength.
The boundary therefore does not only control whether an object crosses, but also preserves how that object must be interpreted once it has crossed.
This allows the model to represent not only valid and invalid responsibility states, but also graded evidentiary conditions relevant for audit and dispute contexts.
Validation Reference
This formulation extends the initial version recorded on 27 April 2026.
The extended validation, including graded evidentiary states, is documented and certified as part of the EVIDE technical record.
Operational reference cases: TGTRACING × EVIDE boundary validation — April 2026, including full validation across rejection, full-strength, and degraded evidentiary states.
External validation surface:
CLARIXO reference-case summary
Explore the protocol infrastructure:
→ The </AI> Protocol
→ From decision to defensible structure: when supervision becomes evidence
→ Read the Public Technical Specification
→ Verify a CWC Code in the Public Registry
→ EVIDE – Evidentiary registry for digital content and decisions
→ CWC Registry Policy
→ Request an Official CWC Verification Code
→ AI Governance Documentation Framework
→ Implementation Guide: verifiable AI supervision
→ Oversight Bias: why human supervision can fail in AI systems
→ Decision Attestation Layer: the missing evidentiary layer in AI governance
→ AI Evidence Officer: proving human supervision in artificial intelligence systems
→ Evidentiary Layer in AI Governance
→ </AI> Protocol FAQ: questions and answers about the framework
Related reading:
→ AI Data Poisoning: The Attack No Antivirus Can Stop
→ Human in the loop: why saying there is human oversight is not enough
→ Real cases: when AI governance fails — and what should have been provable
→ AI Governance: when something has already gone wrong – forensic reconstruction and digital evidence
Work with us:
Understanding the </AI> Protocol
If this is your first time here, this is a quick orientation.
The </AI> Protocol is not a single page. It is a structured system made of distinct layers.
- The </AI> Protocol
The core concept. A public marker that declares human supervision. - Technical Specification
Defines how the system works: tag, structure, verification logic. - Public Registry
Where verification happens. It does not create truth. It allows independent validation. - Registry Policy
Defines the rules: what is required, what is valid, who is responsible. - CWC Code Request
The entry point to obtain an official verification code. - AI Governance Documentation Framework
Explains how to structure supervision, responsibility and traceability. - Implementation Guide
How to apply the protocol in real workflows. - Oversight Bias
Why human supervision often fails in practice. - Decision Attestation Layer
Where decisions become structured, attributable evidence. - AI Evidence Officer
The role responsible for ensuring evidentiary integrity. - Evidentiary Layer
→ Go to the EVIDE evidentiary registry
The core of the system: what makes decisions defensible when challenged.
The protocol is structured around four components: Tag </AI> (public declaration), CWC Code (verifiable identifier), Public Registry (external verification) and FEDIS (independent legal proof with SHA-256 hash and qualified eIDAS timestamp).
In short:
Governance defines how decisions should happen.
The </AI> Protocol makes those decisions provable.
