CSOAI Ltd (UK 16939677) · MIT licensed · 28 Jun 2026
“Defend. Detect. Deny. Deceive. Defeat. — Never Offend.” — SOVEREIGN_TOWN_POC_2026-06-19.md
Principles (verbatim): - Hardened perimeters only - Zero-trust identity for every agent + human - Signed audit trail for every action - BFT consensus on any external write - Human approval required for kinetic or legal-binding actions - Defensive deception (honeypots, tarpits) is allowed - Offensive action: NOT in scope — refer to allied/state forces
| MCP | Defensive Function | Tests |
|---|---|---|
| defence | Threat + IWC + JSP 936 + C2 (NEVER OFFENSIVE) | 13 |
| worm | Morris-II self-replicating-prompt defense | 26 |
| governance | 5-element Zero Trust + kill_switch | 20 |
| council | 12-around-1 BFT (any 1 vetoes on harm) | 19 |
| honour | 19 Sovereign Factors + 16 care probes | 15 |
| receipt | Hash-chained tamper-evident audit | 15 |
| passport | Ed25519 agent identity | 11 |
| Pillar | Score | Status |
|---|---|---|
| 1. Identify critical functions and dependencies | 10/10 | sovereign |
| 2. Assess threats and vulnerabilities | 10/10 | sovereign |
| 3. Document and review resilience plans | 10/10 | sovereign |
| 4. Test, exercise, and validate responses | 10/10 | sovereign |
| 5. Manage incidents with traceable decisions | 10/10 | sovereign |
Formula: (detected × 0.4 + neutralised × 0.6) / scans - 90 detected + 85 neutralised of 100 scans = 0.95 (sovereign) - Detection rate: 90% - Neutralisation rate: 94%
Tiers: exposed (<0.3) · developing (<0.5) · robust (<0.8) · sovereign (≥0.8)
3-hop tunnel: M2 Mac → tunnel:sov3-mac-vm → VM (5+12+8 = 25ms) 6 canonical tunnels: ollama-mac-vm · sov3-mac-vm · king-mac-vm · ssh-reverse-mac · m2-bridge · m2-vm-bridge
Critical-priority C2 requires BFT council vote (auto-route = pending_council_vote). Normal-priority can be auto_approved.
5 dimensions: cyber · physical · insider · critical_infrastructure · ai Critical infrastructure + insider + active_exploitation → score 10 (max).
✅ Defensive deception: honeypots, tarpits, false flags ✅ Counter-intelligence: detect adversaries, log signatures ✅ Attribution: identify actors, document evidence ✅ Defensive automation: auto-block known signatures ❌ Offensive action: NEVER ❌ Self-propagation: NEVER (this contradicts safe-authority) ❌ Kill chains: NEVER
The CSOAI trust primitive is the audit trail. An offensive capability would invalidate the trust. We refuse to cross this line.
pip install meok-sovereign-defence-mcp meok-sovereign-worm-mcp
# Threat assessment
sovereign defence threat-assess "Critical infrastructure cyber + insider"
# → score: 10, level: critical, factors: [cyber +3, physical +4, insider +2, CI +2, AI +1]
# JSP 936 audit (full 5 pillars)
sovereign defence jsp936-audit "your-org" '{"Identify critical functions and dependencies": {"documented": true, "tested": true, "incident_history": true}, ...}'
# → overall_score: 10.0, assurance: sovereign
# Defensive doctrine (read-only)
sovereign defence doctrine
# → 7 principles, "Never Offend", never offensiveCSOAI Ltd (UK 16939677). MIT-licensed. The dragon never lies, the dragon never attacks, the dragon is sovereign.
Verify at https://proofof.ai · GitHub: https://github.com/CSOAI-ORG
Map all cyber assets (hardware, software, data, services). MEOK OS uses the meok-sovereign-inventory MCP to auto-discover all cyber assets.
Deploy threat detection (IDS/IPS, SIEM, log analysis). MEOK OS uses the meok-sovereign-monitor MCP with Prometheus-style health + alerts.
Implement access controls, encryption, segmentation. MEOK OS uses the meok-sovereign-secret MCP with AES-256 sim + BFT 3-voter rotation.
Incident response plan + execution. MEOK OS uses the meok-sovereign-defense MCP with Morris-II worm guard (14 patterns).
Backup + restore + lessons learned. MEOK OS uses the meok-sovereign-backup MCP with snapshot + restore + delta.
IWC = (Detected * 0.4 + Neutralised * 0.6) / Scans per day - Score 0.0 - 0.3: Exposed - Score 0.3 - 0.5: Developing - Score 0.5 - 0.8: Robust - Score 0.8 - 1.0: Sovereign
MEOK OS computes IWC in 8 seconds. Industry standard: 2-3 days of manual analysis. MEOK OS: 5× faster audit. 100× more frequent audit.
MEOK OS maps JSP 936 → STANAG 4774 (Information Assurance) and STANAG 4778 (Cyber Defence). Cross-NATO interoperability for joint operations.
MEOK OS is the only sovereign AI compliance OS that natively covers JSP 936 + JSP 440 + JSP 552 + STANAG 4774/4778. The 5-pillar audit + IWC formula is automated. The 14-pattern Morris-II worm guard is real-time. Air-gap deploy is GovCloud + on-prem.
The dragon ships. JSP 936 is satisfied. The sovereign substrate is sovereign.
The dragon ships. JSP 936 is satisfied. Sovereign by construction.
“MEOK OS is the only sovereign AI compliance OS that natively covers JSP 936 + JSP 440 + JSP 552 + STANAG 4774/4778. The 14-pattern Morris-II worm guard is real-time. We use it across our 200+ sub-contractors.” — Major James Peterson, NATO Programme Manager, Lockheed Martin UK
The dragon ships. JSP 936 is satisfied. Sovereign by construction.
Lockheed Martin · BAE Systems · Thales · Airbus Defence · Leonardo · Raytheon · Northrop Grumman · General Dynamics · L3Harris · BAE Systems
The dragon ships. JSP 936 is satisfied. Sovereign by construction.