Fourteen vendors,
severed and
replaced.

Assesses: origin and transparency of hardware and software design, build, packaging and distribution.

• ▪A documented severance record. Fourteen third-party dependencies identified, replaced and dated — precisely the artifact this objective demands, and the one most bidders cannot produce at any price.
• ▪Self-built distribution. Signed binaries published as content-addressed identifiers and peer-to-peer magnet links, byte-verified across three independent public gateways. No store, no intermediary, no gatekeeper in the chain.
• ▪Own forge and build path. Source hosting, compilation and deployment all run on operator-controlled infrastructure.
• ▪Hardware origin is not European. Compute and accelerators are commodity parts of US design. The software chain is sovereign; the silicon is not, and is not claimed to be.

Assesses: where decisive authority, ownership and financing sit, and how stable that is.

• ▪Clean, unencumbered title. Sole ownership. No investors, no board, no co-founder, no contractor claims, no licensing encumbrance. Chain of title is a one-line answer — which, in a market where the last comparable deal died on ownership grounds, is not a small thing.
• ▪Transferable in full. Because ownership is undivided, an outright assignment moves decisive authority into the acquirer's jurisdiction on closing. No consent chain, no minority holdout.
• ▪Not EU-owned today. This objective is satisfied by the transaction itself rather than by the asset — which is exactly why a European acquirer extracts more value here than a non-European one.

Assesses: whether EU actors can independently run, support and develop the service with no non-EU involvement.

• ▪Zero-downtime deployment on owned iron. Blue/green cutover with health-gated traffic swap, currently carrying four production applications.
• ▪Operates disconnected. Demonstrated serving full interface, API and authentication with both the compute host and the public front door unreachable — the scenario every continuity plan describes and almost none has tested.
• ▪Currently one operator. Independent operation is achievable and documented, but not yet staffed by a second party. This is exactly what a bounded transition period exists to resolve, and it is scoped as such rather than deferred.

Assesses: openness of the stack — standards, open licensing, auditability, and freedom from vendor lock-in.

• ▪No proprietary runtime anywhere in the path. Every replacement is an open, self-hostable component under operator control.
• ▪Lock-in removed by construction, not by assertion. The platform demonstrably runs with its vendors removed — that is how each replacement was validated in the first place.
• ▪Auditable by design. A 217-entry engineering corpus records every incorrect assumption made during the build, the check that would have caught it, and the standing rule that replaced it. Few codebases of any size can hand a buyer their own failure history.
• ▪Reversible cutovers. Retired paths are ghosted rather than deleted, so a migration can be rolled back under load instead of being a one-way door.

Assesses: cryptographic control over data, processing location, and independence of AI capabilities.

• ▪Inference never leaves the building. Twelve specialist agents run on locally owned accelerators. No prompt, document or user record is transmitted to a model provider — the constraint no hosted API satisfies at any price, for any customer, under any contract.
• ▪Own search substrate. Research queries route through a self-hosted metasearch node, so even an agent's lookups disclose nothing to a third party.
• ▪Own identity and key material. Passkey and code-based authentication replaced a hosted identity vendor. No custodial third party holds user credentials.
• ▪One managed database remains off-box. Disclosed in full at §05. It is the single largest open engineering item and it is named here rather than discovered later.

Assesses: exposure to foreign laws with extraterritorial reach, and whether non-EU authorities can compel access.

• ▪Almost no third party left to compel. With hosting, mail, identity, storage, search and inference all self-operated, the list of external processors an authority could serve is close to empty. Sovereignty here is a consequence of architecture, not of contract language.
• ▪Data resides where the hardware resides. Processing location is a physical fact about an owned machine rather than a regional promise in a service agreement.
• ▪Resolved on transfer. As with SOV-1, extraterritorial exposure is a function of who owns the asset. An EU assignment removes it.

Assesses: EU-controlled security operations, formal certifications, and GDPR / NIS2 / DORA alignment with audit rights.

• ▪Security operations are in-house. No outsourced monitoring, no third-party agent holding privileged access to production.
• ▪No formal certifications held. ISO 27001, NIS2 and DORA attestation work has not been undertaken. This is the objective with the furthest to travel, and an acquirer with an existing management system will close it faster than this asset could alone.

Assesses: energy efficiency, renewable sourcing, circular practice and reporting transparency.

• ▪Genuine circularity. Production inference runs on prior-generation workstation accelerators deliberately kept in service — reuse rather than procurement, which is what the objective actually rewards.
• ▪A very small footprint. The entire production estate is one workstation, against a third-party cloud floor of roughly $20 per month.
• ▪No formal reporting. Efficiency and sourcing are not yet measured or published to any standard.