The sovereign defence AI
that's 94-98% cheaper than Palantir.

EU AI Act enforcement in 26 days.

Every UK + EU + AUKUS-Pillar-2 prime + Tier-1 supplier is being asked by their board:

• Article 9 RMS — 7-step risk management
• Article 14 human oversight — 9 layers, kill-switch <2s
• Article 26 deployer obligations — 10 categories + DPIA
• Article 47/48/49 CE marking + declaration + register
• Article 50/52 transparency + C2PA watermarking
• UK JSP 936 / DSEC, NATO Article 13, AUKUS Pillar 2

DEFONEOS: 13 model classes in one organic loop.

The sovereign substrate that runs:

• Mamba-2 SSM (16-dim state, ~3,000 tok/s)
• 64-expert MoE + Standard Attention
• qwen3:0.6b (Mac) + gemma3:4b (GCP UK)
• 6 trained NNs (1,793 samples)
• HotStuff BFT 33-agent council (200ms finality)
• Ed25519 SIGIL chain + OTS Bitcoin anchor
• Care Membrane (847 signals × 23 categories)
• OSCAL attestor (236 frameworks)
• bridge_think MCP (#116, 4 profiles)

5 structural moats.

1. Audit-receipt moat — Ed25519 SIGIL chain baked in from day one. Palantir/Anduril can't retrofit.
2. Sovereignty moat — 100% UK-only compute. Charter Article 0 binding. ISO fee-for-service only.
3. Care-floor moat — 7 protocol-enforced red lines (no kinetic, no surveillance, no civilian harm, etc.). Primes can procure us where they cannot procure Palantir.
4. Open-source distribution moat — 30 MCPs on PyPI + 15 public repos. Procurement cost zero. Once adopted, Crown is in the catalogue.
5. Compliance-surface moat — 236 framework crosswalks. Primes need us, not a vendor.

5-year TCO: £285K DEFONEOS

vs the incumbents:

94-98% cheaper

9 oversight layers + 7 red lines.

Every sovereign action goes through:

• HIC / HITL / HOTL oversight
• Kill-switch (<2s)
• External audit hook (mTLS-pinned)
• 5-tier explanation framework
• Subject rights matrix

7 protocol-enforced red lines. No human approval, no BFT vote, no override, no failure mode can bypass.

Forecast (mid-case):

Driven by: 7 primes in evaluation, 3 regulators in cooperation, 19 government-dept sovereign pitches, AUKUS Pillar 2 £3.5B / 5yr.

225+ live pages. 30 MCPs. 13 model classes.

The substrate is not a deck. It is running:

• 225+ production pages at csoai-static-deploy2.vercel.app
• 30/30 MCPs in production on PyPI
• 13 model classes in one organic loop
• SIGIL chain emitting at 1Hz (86,400/day)
• Audit-trail access protocol live (regulator-cooperation-grade)
• OSCAL attestor live, JSON+XML export
• Bridge_think MCP (left/right bilateral cognition)

£45-90M Series A · 12-week close

Charter Article 0 binds CSOAI indefinitely.

ISO fee-for-service only. No equity, no board seats, no success fees from institutions we certify. (S&P paid $1.375B and Moody's paid $864M in settlements for this exact conflict.)

This term sheet does not override Charter Article 0. Amendment requires 33-of-33 BFT + 5 human sigs + 14-day window + 90% supermajority. Has never been triggered.

Board composition (post-close).

• 3 of 5 (60%) — Founder (lifetime, Charter Article 0 binding)
• 1 of 5 (20%) — Series A lead (3-year term)
• 1 of 5 (20%) — Independent UK / AUKUS-aligned (e.g. ex-MI6, ex-NSA, ex-DSTO)

Plus: BFT 23/33 sovereign council as a layer below the corporate board for Charter Article 0 issues.

4 viable exit paths.

Lead-ticket next steps.

1. Email series-a@csoai.org with "Series A Lead" in subject.
2. Receive 18-page investor brief (within 24h) + 30-min scoping call (within 7 days).
3. Data room access (technical + financial) + 4-week exclusivity (if lead is appropriate).
4. BFT 23/33 council vote on the lead (Charter Article 0 governance compliance).
5. Final term sheet negotiation + sign + close (target Q4 2026).