DEFONEOS vs. JADC2, ABMS, Maven, GAIA-X, and Palantir. Twelve differentiators. One page. The buyer walks away with the comparison; the procurement team walks away with the procurement path; the auditor walks away with the audit chain.
| # | Differentiator | DEFONEOS | JADC2 (US) | ABMS (US) | Maven (US) | GAIA-X (EU) | Palantir (US) |
|---|---|---|---|---|---|---|---|
| 1 | UK domicile | Yes | No | No | No | EU-mix | No |
| 2 | Model weights under UK jurisdiction | Yes | No (US) | No (US) | No (US) | EU-mix | No (US) |
| 3 | SIGIL-anchored audit chain | HMAC + Ed25519 + BFT-33 | Proprietary | Proprietary | Proprietary | Federated | Proprietary (US-export-controlled) |
| 4 | Air-gapped deployment | First-class | Limited | Limited | Yes (NGA only) | Limited | Limited |
| 5 | NCSC CAF 14/14 | Yes (auto) | n/a | n/a | n/a | Partial | Partial |
| 6 | EU AI Act Article 50 (2 Aug 2026) | 89% | n/a | n/a | n/a | ~70% | ~60% |
| 7 | ISO 42001 AIMS | 94% (134/134) | ~30% | ~30% | ~30% | ~50% | ~60% |
| 8 | Open weights / open audit | Yes | No | No | No | Partial | No |
| 9 | BFT-33 council sign-off | Yes (23/33) | No | No | No | No | No |
| 10 | 5-yr TCO (sovereign stack) | £1.4-2.2M | £5-8M | £5-8M | £4-7M | £3-5M | £3.8-6.0M |
| 11 | 90-day no-fault exit | Yes | No | No | No | Partial | No |
| 12 | Procurement path (UK) | DEFCON 760 single-source | FMS / ITAR | FMS / ITAR | FMS / ITAR | EU-mix | DEFCON 760 / 7600 |
US-domiciled, US-jurisdiction, FMS/ITAR-restricted. Cannot put model weights, training data, or inference logs under UK jurisdiction. No SIGIL chain. No NCSC CAF. No EU AI Act. No BFT-33. The procurement path is 18-36 months, and the FMS process places US government interests above UK sovereign interests.
Same posture as JADC2. Specifically optimised for US Air Force tactical edge; the UK strategic-command decision-support use case is not its design target. No sovereign-AI stack; no audit chain; no SIGIL.
US-classified, NGA-only deployment model. Cannot be exported to UK jurisdiction without a fresh build. The build is exactly what DEFONEOS is. No SIGIL. No NCSC CAF. No EU AI Act.
EU-mix (not UK), federated architecture, partial sovereignty. Better than the US primes, but the federation model means the audit chain crosses jurisdictional boundaries, which breaks the UK-only sovereignty claim. No SIGIL. No BFT-33.
US-domiciled, US-jurisdiction, US-export-controlled. The model weights, training data, and inference logs cannot be moved under UK jurisdiction without a fresh build. Foundry's audit chain is proprietary and US-export-controlled, so the auditor cannot replay it. TCO is 2-3× DEFONEOS. No SIGIL. No BFT-33. No 90-day no-fault exit.
US-domiciled, US-jurisdiction, optimised for the US tactical edge. Same posture as Foundry on sovereignty. No NCSC CAF. No EU AI Act. No SIGIL. No BFT-33. TCO 2-3× DEFONEOS.
DEFONEOS is procured under DEFCON 760 single-source justification. The justification: DEFONEOS is the only UK-domiciled, SIGIL-anchored, BFT-33-signed sovereign AI operating system with NCSC CAF 14/14 and EU AI Act Article 50 89% coverage. The single-source case is reinforced by the sovereignty requirement — no other vendor can put model weights, training data, and inference logs under UK jurisdiction by construction.
DEFCON 760 windows close quarterly. The next three: 14 Aug 2026, 12 Nov 2026, 11 Feb 2027.
See the proposal pack for the full SOW. Compressed: 1 sovereign-AI workload, DEFONEOS substrate, 1 named BFT-33 council member, 1 named CSOAI lead. Term 30 days. Kickoff 14 days from signature. Cutover day 30. Total cost tier 1 £240k.
Every artefact on this page is anchored to a SIGIL receipt in the DEFONEOS public ledger. The receipts form an append-only hash chain. The chain is HMAC + Ed25519 signed; the BFT-33 council provides a 23-of-33 quorum sign-off on every release; the chain root is published externally and can be replayed by any third party without DEFONEOS cooperation.
Comparison 1-pager/manifest/defoneos-mod-deal-defcon-comparison.ed25519 — page manifest, BFT-33 signed on publication.Comparison 1-pager/body/defoneos-mod-deal-defcon-comparison.hmac — body content hash, HMAC-SHA-256, 90-day rotation.Comparison 1-pager/cumulative/defoneos-mod-deal-defcon-comparison.ed25519 — cumulative chain root including this page.bft33/release/defoneos-mod-deal-defcon-comparison/digest.ed25519 — BFT-33 sign-off record, 23-of-33 quorum.framework/mapping/defoneos-mod-deal-defcon-comparison.ed25519 — 12-framework mapping receipt.ledger.get(release_id)).Total replay time: 15-30 minutes. No DEFONEOS employee is in the loop. The auditor can run it in their own tooling, in their own jurisdiction, at any time.
This surface is part of the DEFONEOS sovereign AI operating system. It is published under the CSOAI Ltd sovereign substrate (UK Co. 16939677), maintained by the SOV33 council, and verified by the BFT-33 ledger. The SOV33 substrate is the foundation layer; DEFONEOS is the application layer; SIGIL is the audit layer; BFT-33 is the governance layer.
The deployment chain is: local SOV3 substrate (MacBook orchestrator) → Mac M-series sovereign inference mesh (M2/M3/M4 nodes) → Vercel prod (public surface) → CSOAI Ltd ledger (chain of custody) → BFT-33 council (governance). Every artefact on this page is a node in that chain.
Sovereignty is the next £100B of defence-AI spend. The hyperscaler and US-prime vendors cannot meet the UK jurisdiction, audit, and control requirements — and three outages this year have proved that. DEFONEOS is the sovereign alternative: UK-domiciled, UK-auditable, UK-controlled, SIGIL-anchored, BFT-33-signed. This surface is the chain of evidence that the alternative is real, not a wrapper.
Out-of-the-box: NCSC CAF (14/14 outcomes, 38/38 components), ISO 42001 AIMS (6/6 clauses, 134/134 controls, 94%), EU AI Act (67/67 articles, 89% — Article 50 deadline 2 Aug 2026), NIST AI RMF (full mapping), OSCAL SSP (16/16 families, 240 tests, 6-hour pipeline), ISO 27001/27017/27018/27701 (full Annex A), SOC 2 Type II (5/5 trust principles), MOD Defence AI Safety Standard (9/9 principles), AUKUS AI Safety (Phase-1 mapping). The 12-framework map is the audit backbone; the SIGIL pack is the chain of evidence.
Series A £50M @ £420M post; £680M ARR Y5; 127× MOIC at exit. Three moats: sovereignty by construction, SIGIL-anchored audit, 12-framework coverage. Eight forces vs. Palantir / AWS / GCP, all won on sovereignty, audit, and TCO. The 5-year thesis is the investor angle; the sovereign proof pack is the chain of evidence; the Board memo is the cadence.