v3.0.2 Spec v3.0 · Reference implementation v3.0.2 (current); v3.0.0 shipped 2026-06-01. RFC 0002 Accepted and implemented. Version history →
Open Source  ·  Apache 2.0 / CC BY 4.0  ·  v3.0.2

Sovereign, auditable, neuro-symbolic AI compliance

An open framework for regulated industries — runnable today.

Every verdict carries verifiable provenance: a signed LLM transcript, Knowledge Graph citations, and an independent symbolic verifier's per-assertion result.

Read the docs → Quick start
0
Axioms
0
Pillars
0
Verdicts
0
Phases
Apache 2.0 / CC BY 4.0 Spec v3.0 · Impl v3.0.2 Sovereign by default — Ollama runtime 60+ conformance tests, runnable today
ski-runtime  ·  verdict stream ILLUSTRATIVE
CLEAR FLAG DISCRETIONARY NULL_UNMAPPED NULL_STALE
Illustrative verdict stream — sample tags and verdicts, not live operational data.
scroll
The verdict envelope

A real verdict envelope. Every evaluation produces exactly this shape — categorical verdict, KG citations, formalizable assertions, the Symbolic Verifier's cross-check result, and the six hash anchors needed to replay it. Hover any field to see the audit guarantee it carries.

V3VerdictEnvelope · spec v3.0 §4.2
{  "verdict" "FLAG"  "reasoning" "SO2 discharge at 142 ppm exceeds the §60.2(a) cap of 100 ppm."  "kg_citations"
    
      "node_id" "energy.so2.lte_100ppm"
      "role" "obligation"
      "source_document" "40 CFR 60.2(a)"
      "source_clause" "Subpart A — General Provisions"
    
    "formalizable_assertions"
    
      "obligation_id" "energy.so2.lte_100ppm"
      "metric" "so2_ppm"
      "operator" "must_not_exceed"
      "observed" 142
      "expected" 100
      "unit" "ppm"
      "satisfied" false
    
    "verifier_result"
    "status" "AGREED"
    "per_assertion"
      
        "obligation_id" "energy.so2.lte_100ppm"
        "status" "AGREED"
        "verifier_observed" 142
        "verifier_expected" 100
      
    
    "model_provenance"
    "model_weight_hash" "sha256:7c2d1f8a…"
    "kg_version_hash" "sha256:eb21125f…"
    "prompt_template_id" "v3-evaluator-jurisdictional-1"
    "prompt_template_hash" "sha256:9a8b1e23…"
    "decoder_seed" 0
    "structured_grammar_hash" "sha256:1f4c8b9d…"
    "transcript_ref" "ledger:tenant.demo/seq:00042"
}
Production hardening at speed

v3.0.0 shipped 2026-06-01. v3.0.1 (same day) and v3.0.2 (next day) closed two real-world deployment gaps an external tester reported — the runtime now self-heals an out-of-date ledger schema on startup, and the fresh-deploy path is pinned by a regression conformance test. The pace from first production-target release to tested end-to-end in a deployer's hands was 24 hours.

The Problem

Six reasons regulated industries won't deploy AI — and how SKI resolves each one

These are not hesitations. They are the reasons AI adoption has stalled in core compliance functions across critical infrastructure. SKI was designed from first principles to address each one.

01
Loss of Sovereignty
If our operational data flows through a third-party cloud AI, we lose control of our most sensitive asset.
SKI resolves this The inference engine runs entirely within the organisation's sovereign boundary. No operational data leaves at runtime. Air-gap capable by design.
02
Non-Determinism
AI gives different answers to the same question. That's fundamentally incompatible with audit and regulatory compliance.
SKI resolves this Every verdict carries verifiable provenance: a signed LLM transcript, KG citations, model-weight and KG-version hashes, and an independent Symbolic Verifier's per-assertion result. An auditor can replay any verdict from the recorded transcript and reach the same envelope — defensibility is reconstructible, not bit-identical.
03
Liability Opacity
When an AI flags a compliance issue (or misses one), we can't explain the basis of the decision to a regulator.
SKI resolves this Every verdict traces to a specific Knowledge Graph node and a named policy clause in a source regulatory document. Every verdict also carries a signed LLM transcript so an auditor can reconstruct the model's reasoning step by step. No black boxes — ever.
04
Unsafe Intervention
An AI that can control or modify operational systems introduces catastrophic risk. We cannot put AI in the loop on safety-critical infrastructure.
SKI resolves this SKI is a passive read-only observer. It monitors and reports. It has zero control path to operational systems. Primary operations continue uninterrupted if SKI goes offline.
05
Rule Drift
Regulations change. We can't guarantee the AI is evaluating against the current rules, not last year's.
SKI resolves this The Knowledge Graph is a living, versioned artefact with a governed update pipeline. Every change requires extraction, human validation, cryptographic signing, and deployment. The ledger records which graph version produced each verdict.
06
Accountability Gap
If the AI is making compliance decisions, who is accountable? The regulator will ask, and we won't have an answer.
SKI resolves this Human reviewers validate every Knowledge Graph rule before production. Human authority is preserved on all escalations. The AI augments judgement — it does not replace it.
0
Provenance + durability conformance tests passing
30 provenance + 30 durability
0
Releases in 24 hours
v3.0.0 → v3.0.1 → v3.0.2 — schema gaps caught and fixed in one day
0
Outbound network calls during evaluation
Sovereignty axiom is architectural, not policy
0
Verdict types
Closed taxonomy. No confidence scores, no free-form text.
Framework Overview

SKI at a Glance

Three governing axioms. Three structural pillars. Five categorical verdicts. One sovereign boundary. Click any element to open its full specification in the docs.

3 Governing Axioms
A1 Proximity Inference at or near the data source. No remote endpoints. A2 Verifiable Provenance Every verdict carries a signed LLM transcript, KG citations, and the Symbolic Verifier's per-assertion result. Replayable by any auditor. A3 Passive Oversight Read-only sidecar. Zero control path to operations.
govern
3 Structural Pillars
S
Sovereignty
On-premise edge compute Air-gap capable Hardened environment No cloud dependency
K
Knowledge Maps
LLM extraction (kg-extractor) Schema validation (kg-validator) Ed25519 signing Typed v3 obligation graph
I
Intelligence
KG-grounded local LLM Symbolic Verifier Signed LLM transcripts Agreement monitor
evaluated by
Evaluation Architecture
Phase 1 — Offline
Knowledge Graph Compilation
Extract → Validate → Sign → Package
Outside sovereign boundary
one-way
transfer
Phase 2 — Runtime
On-Premise Evaluation
Listen → Tag → Evaluate → Log
Inside sovereign boundary
LLM Evaluator KG-grounded local LLM — temperature 0, structured generation, signed transcript Symbolic Verifier Mechanically cross-checks every formalizable assertion — runs on every verdict
produces
5 Verdict Types
CLEAR All rules pass — logged silently FLAG Rule breached — operational owner alerted NULL_UNMAPPED No graph node — KG coverage gap NULL_STALE Silent sensor — infrastructure failure DISCRETIONARY Ambiguous language — human reviewer
Explore the full specification →
Sector Applications

Built for industries where compliance failures cost lives, licences, or billions

SKI was architected for environments where regulators audit every decision, data sovereignty is non-negotiable, and system failure is not an option.

01
Oil & Gas
Applicable regulatory contexts
Provincial energy regulators · Environmental protection statutes · Operational safety codes
Continuous wellhead pressure, flow rate, and emissions monitoring against permit conditions. Detection of threshold breaches before reporting deadlines. SCADA integration via read-only OPC-UA sidecar.
SCADA integrationAir-gap capableWellhead monitoring
02
Financial Services
Applicable regulatory contexts
Model risk supervisory guidance · Operational resilience regimes · Conduct and reporting obligations
Model risk governance and trading compliance monitoring with full audit trail. Every algorithmic decision traced to a named policy clause for regulatory examination. No cloud exposure for sensitive position data.
Model governanceRegulatory reportingTrade surveillance
03
Defence
Applicable regulatory contexts
Controlled-information regimes · Defence cybersecurity maturity frameworks · Classified handling controls
Classification handling compliance and information access monitoring in air-gapped environments. Cryptographic integrity verification on every compliance verdict with zero external network dependency. Full audit trail for security reviews.
Air-gapped deploymentClassification controlsZero cloud
04
Energy & Utilities
Applicable regulatory contexts
Bulk electric system reliability standards · Industrial control system security frameworks · Energy regulator directives
Grid operations compliance monitoring with sub-100ms target verdict latency on dedicated edge hardware. Reliability standards enforcement across distributed substations with on-premise edge nodes per site. No central aggregation of sensitive operational data.
Grid operationsEdge deploymentDistributed sites
05
Mining & Resources
Applicable regulatory contexts
Environmental protection statutes · Tailings management standards · Worker safety regulations
Environmental monitoring compliance for tailings, water treatment, and air quality against permit conditions. Continuous tracking of reportable events with tamper-evident ledger available directly to regulators.
Environmental monitoringTailings complianceAudit ledger
06
Critical Manufacturing
Applicable regulatory contexts
Electronic records and signature regulations · Quality management standards · Good manufacturing practice
Manufacturing process compliance and deviation detection in pharmaceutical and medical device production. Complete electronic records with validated audit trails. Passive monitoring with zero impact on batch processes.
Process deviationElectronic recordsGMP compliance
Sector details & crosswalk roadmap → How SKI compares →
Runnable today — v3.0.2

Clone the repo and run all 60+ conformance tests in minutes

The reference implementation runs entirely on-premise on Ollama — no cloud API key required. The full quick-start, conformance model, architecture, and governance specification live in the documentation.

git clone https://github.com/kpifinity/ski-framework.git
cd ski-framework && pip install -r requirements-dev.txt
pytest conformance/ -m "provenance or durability" -v
Open the Quick Start → View on GitHub ↗