Platform

Four layers, four questions

Together they control what AI is permitted to decide or execute, and produce independently verifiable evidence of what occurred, under which rules and using which evidence.

Talk to us about a pilot See the implementations

evidence-pack.json accept

// returned by the lane, signed before it is returned
{
  "verdict": "accept",
  "lane": "dela.evidence-change",
  "reason": "material change in coordinates; 3 dependent analyses affected",
  "inputs": {
    "current":   "sha256:9f2c…a71d",
    "candidate": "sha256:4b80…c2e9"
  },
  "impact": {
    "affected":  3,
    "preserved": 41,
    "replayed":  3,
    "held_for_review": 1
  },
  "deterministic": true,
  "chain": { "index": 18274, "prev": "sha256:0c15…8fa2" },
  "sig": "ed25519:7d41c0…9b3e"
}

Verify with the published key. No source access required. Replayable

A verdict as returned. Categorical, not probabilistic, with the dependency impact and a signature checkable against the published key.

On this page The four layers Operating model Integration Deployment

Four questions a regulated organisation has to answer

Existing systems record versions and produce governance reports. They do not answer these, quickly, with evidence a third party can check.

DTL

Deterministic Taxonomy Lanes

Is this output or decision allowed?

DTL routes an output or decision into a defined verification lane, applies machine-readable rules, and returns a categorical result with the reasoning recorded. It does not score confidence. It decides, or it declines to decide.

10 implementations · AU provisional 2026905289

DAX

Deterministic Action Execution

Is this physical action allowed to happen?

DAX places a permission boundary between a decision and an action that affects hardware, machinery, infrastructure or another physical system. It is fail-closed: if the check cannot complete, the action does not proceed.

4 implementations · AU provisional 2026905506

DCLA

Deterministic Control Lineage Architecture

The governing rules changed. What does that affect?

DCLA tracks changes to regulations, policies, controls and operating requirements, then identifies the systems, decisions, evidence and previous approvals that may need reassessment. Affected work is held rather than allowed to continue on a superseded rule.

3 implementations · AU provisional 2026907053

DELA

Deterministic Evidence Lineage Architecture

The underlying evidence changed. What does that affect?

DELA detects changes in scientific, operational or other source evidence, traces the affected dependencies, preserves the work that remains valid, selectively replays reproducible operations, and routes non-reproducible work to qualified human review.

5 implementations · AU provisional 2026907052

Operating model

How the layers connect

DCLA

A regulation, policy or internal control changes. The affected systems, decisions, evidence and prior approvals are identified and held.

DELA

Source evidence changes. Affected work is separated from work that remains valid, reproducible operations are replayed, and the rest is routed to qualified review.

DTL

The resulting output or decision is checked against the applicable control, returning accept, block, review, abstain or pending evidence.

DAX

Where the decision would drive a physical system, a fail-closed permission boundary sits between the decision and the action.

Evidence

A signed, tamper-evident record states what was checked, what changed, what was preserved, who reviewed what and what was allowed.

Integration

It surrounds your systems. It does not replace them.

Receives

AI model outputs and workflow decisions
Scientific and operational source evidence
Control and policy states
Proposed physical actions

Processes

Normalisation into machine-readable components
Control and evidence change analysis
Deterministic lane execution
Action permission where execution is physical

Returns

A categorical verdict, or an explicit abstention
The affected set and the preserved set
A qualified review queue
A signed evidence pack

What EcoKure is not

Not a general-purpose AI model. Not a replacement for your governance, risk and compliance platform, your quality management system or your scientific tooling. Not a certification. It sits around the systems you already run and produces checkable evidence about them.

Deployment

Enterprise deployment

Concern	Position
Deployment model	Cloud, on your own infrastructure, or hybrid. The verification lanes are deterministic and do not require an external inference call.
Data residency	Lanes can run entirely inside your network. Where they do, no output, evidence or prompt leaves it.
Tenancy	Multi-tenant with isolation, or single tenant on your infrastructure.
Evidence signing	Ed25519 signatures with a tamper-evident chain. A third party can verify a pack without our source and without trusting us.
Availability behaviour	Fail-closed. An unreachable gate is recorded as unavailable, never as a pass. This is the property auditors ask about first.
Integration	A metered API, with connectors defined per workflow during a pilot rather than assumed in advance.
Certification	Controls are built to be auditable. EcoKure holds no certification today, and says so rather than implying otherwise.

Start with one workflow

The useful first conversation is thirty minutes on a workflow where a change in rules or evidence has already cost you work, so both sides can tell quickly whether this is worth pursuing.

Talk to us Regulated life sciences