EVIDE does not judge decisions. It makes them defensible.
The gap that governance cannot close
Every AI governance framework defines how decisions should happen. Who reviews, who approves, what thresholds apply, what taxonomy is used.
But there is a structural gap between defining a process and proving that process was followed.
When a decision is contested — in an audit, a dispute, a regulatory inquiry — the question is never abstract:
“Was this decision made within a defined structure, or was it improvised?”
A system log cannot answer that question. It can show that something happened. It cannot show whether that something was anchored to a defined taxonomy, a declared threshold, a verifiable authority.
That is the gap EVIDE v1.7 addresses.
What changed in EVIDE JSON 1.7
EVIDE JSON 1.7 introduces a new optional field inside the evidentiary record:
"intervention": {
"classification_context": {
"taxonomy_reference": "https://...",
"threshold_reference": "https://...",
"threshold_status": "met"
}
}
This field does not validate decisions.
It exposes whether a decision was made within a defined structure.
The threshold_status field accepts four values:
met— the threshold was defined and respectednot_met— the threshold was defined but not metunknown— a threshold existed but its status cannot be determinednot_defined— no upstream threshold existed at the time of the decision
Invalid values are rejected at ingestion level with a 422 error. There is no soft interpretation of structure. Either it is defined, or it is rejected.
Three concepts that define v1.7
These three concepts redefine how decisions can be understood, defended, and audited.
1. Coordinated position
Before v1.7, an EVIDE record was a photograph of a moment: what happened, who decided, when.
With classification_context, the record becomes a coordinated position — a point that exists in relation to a defined reference grid. The decision is no longer isolated. It is anchored to:
- a taxonomy (the classification system in use)
- a threshold (the admissibility condition applied)
- a version (the exact structure valid at that moment)
A coordinate without a grid is just a point in space. A coordinate with a grid is a verifiable position.
2. Procedural shield
In a legal or compliance context, the traditional defense of a decision looks like this:
“Here is the log. The system recorded that a human approved it.”
This is a weak defense. It proves that something happened. It does not prove that the decision was made within a defined, documented structure.
With v1.7, the defense changes:
“The decision was made following Taxonomy X, Version Y, which was our documented standard at that moment. The threshold was defined and the record shows whether it was met, not met, or not applicable.”
This is not a defense of the outcome. It is a demonstration of diligence in the process. That distinction matters enormously in audit and dispute scenarios.
3. Organizational maturity gauge
The not_defined value is not a failure state. It is a diagnostic signal.
If a significant portion of records show threshold_status: not_defined, that is not evidence that supervisors made wrong decisions. It is evidence that the upstream governance layer has a gap — that the policy is incomplete, not that the people are negligent.
This shifts the frame of accountability in a critical way:
| threshold_status | What it reveals |
|---|---|
met |
Structured governance, decision within defined parameters |
not_met |
Governance exists, exception was flagged and documented |
unknown |
Threshold existed but state is ambiguous — grey zone documented |
not_defined |
No upstream threshold — policy gap, not individual failure |
EVIDE does not assign blame. It makes the structure — or its absence — observable over time.
The separation that makes this possible
A key architectural decision in v1.7 is the separation between two distinct reference types:
- taxonomy_reference — points to the internal classification system the organization adopted. This is an editorial choice: how the organization has chosen to categorize decisions.
- threshold_reference — points to an external rule, regulatory parameter, or admissibility condition. This is often a normative constraint: what the law, regulator, or standard requires.
This separation is not a technical detail. It is what allows governance and regulation to coexist without being confused.
These two references tend to live in different logical domains:
taxonomy_reference: internal risk classification system
threshold_reference: EBA or national regulatory parameter
taxonomy_reference: internal evaluation criteria
threshold_reference: non-discrimination requirements under applicable law
taxonomy_reference: internal clinical protocol
threshold_reference: ministerial guideline or ISO standard
taxonomy_reference: firm-level decision classification
threshold_reference: regulatory or contractual obligation
When both references point to the same source, that is itself a diagnostic signal: the organization has not yet developed its own interpretive layer. That level of governance maturity becomes visible in the record.
EVIDE does not impose the separation. It makes it observable.
What changes for organizations
Decisions become traceable
Every decision carries a reference to the structure it was made within, not just a timestamp of when it happened.
Responsibility becomes attributable
The human supervisor is linked to a specific taxonomy and threshold, not just to an approval action.
Structure becomes visible
Whether governance is mature or incomplete, EVIDE makes that condition observable — not as judgment, but as evidence.
Risk becomes measurable
Aggregating records over time reveals patterns: where thresholds are missing, where policy gaps exist, where governance needs to evolve.
What this means for compliance and legal professionals
For Compliance Officers, DPOs, and legal counsel, the practical value of v1.7 can be stated simply:
“The supervisor approved the decision.” → Log shows an event.After v1.7
“The supervisor approved the decision within Taxonomy X, against Threshold Y, and the record shows the threshold was met.” → Record shows a coordinated position within a documented structure.
The decision is not only recorded. It is anchored to the exact version of the structure that produced it.
This is what turns supervision into evidence.
And evidence into something that can be defended.
Explore the EVIDE JSON 1.7 schema
Explore real use cases
How EVIDE connects to the </AI> Protocol
EVIDE is the evidentiary layer of the </AI> Protocol. It does not replace governance. It does not define who holds authority or how decisions should be structured upstream.
It operates downstream — at the point where a decision has been made and needs to become externally accountable.
- Governance defines how decisions should happen.
- The </AI> Protocol makes supervision verifiable.
- EVIDE anchors the decision state independently, outside the system that produced it.
- EVIDE v1.7 additionally exposes whether that decision was made within a defined structure.
Together, they move an organization from declared supervision to demonstrable, defensible supervision.
→ 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 FAQRelated reading:
→ Human in the loop: why saying there is human oversight is not enough
→ Real cases: when AI governance fails
→ AI Governance: forensic reconstruction and digital evidence
Work with us:
This documentation constitutes a verifiable, timestamped record of its structure, concepts, and implementation.
