Private is not the same as secure

Security controls with a named owner on both sides.

Local hosting can reduce external disclosure, but it also moves responsibility onto the organisation. The handover should make control ownership visible rather than imply the box is secure forever.

Decision in one minute

Buy only when the customer has a technical owner for identity, network, backup, updates and incident response - or contracts those duties separately.

Diagram showing an approved request, a local service, an approved store and a policy-controlled data path
Diagram showing an approved request, a local service, an approved store and a policy-controlled data path
A private deployment starts with the permitted data path, access policy and logging boundary. Original explanatory plate. It sets out a decision method, not a measured result.
A private deployment starts with the permitted data path, access policy and logging boundary.

Operational ownership

The handover should make responsibility easier to see.

01 / Owner
Name who controls updates, access, recovery and incidents.
02 / Record
Make versions, tests and handover evidence visible.
03 / Exit
Keep portability and supplier boundaries written down.
Diagram showing approved documents moving through a searchable index to an answer with a source citation
Diagram showing approved documents moving through a searchable index to an answer with a source citation
A retrieval workflow should connect each useful answer to approved source material and defined refusal behaviour. Original explanatory plate. It sets out a decision method, not a measured result.
A retrieval workflow should connect each useful answer to approved source material and defined refusal behaviour. Original explanatory plate. It sets out a decision method, not a measured result.

The proposed technical baseline

Named administrator access, SSH hardening, host firewall, controlled network exposure, TLS where appropriate, secrets transfer, version records, GPU/system metrics and a defined log-retention setting form the starting point.

The exact controls depend on whether the system is connected, restricted, hybrid or offline.

  • Identity and least privilege
  • Network segmentation
  • Encrypted backups where ordered
  • Documented recovery and update state

Change record

A controlled change leaves an evidence trail

  1. Request Reason, owner and affected service.
  2. Test Focused acceptance and rollback plan.
  3. Approve Named person accepts the change.
  4. Record Version, date and result retained.
Silent model, runtime or access changes make later assurance harder. Decision aid for: The proposed technical baseline

Look behind the claim at the operating record.

Management, recovery, evidence and physical serviceability make responsibilities easier to inspect at handover.

GPU server remote management dashboard with system status, access logs and sensor monitoring panels
GPU server remote management dashboard with system status, access logs and sensor monitoring panels
Supplier screenshot of the platform management interface. The final management features and access policy depend on the ordered system. OEM supplier reference image. Written reuse permission pending.
Supplier screenshot of the platform management interface. The final management features and access policy depend on the ordered system. OEM supplier reference image. Written reuse permission pending.
Diagram of an evidence pack containing an asset schedule, burn-in record, health readings, workload test and admin guide
Diagram of an evidence pack containing an asset schedule, burn-in record, health readings, workload test and admin guide
A credible handover records the supplied assets, checks, operating evidence, instructions and unresolved items. Original explanatory plate. It sets out a decision method, not a measured result.
A credible handover records the supplied assets, checks, operating evidence, instructions and unresolved items. Original explanatory plate. It sets out a decision method, not a measured result.

Customer responsibilities remain real

The customer owns user lifecycle, acceptable use, data and model approval, physical access, retention, data-subject handling, incident response and daily availability unless a separate agreement changes that.

A RACI is part of a serious deployment, especially where a marketplace mode or cloud fallback is considered.

Recovery boundary

Recovery spans equipment, configuration and business data

  1. Equipment Hardware fault and warranty route.
  2. System Build record, settings and secrets.
  3. Data Backup, retention and restore authority.
  4. Service Fallback, incident and return to use.
A backup claim is incomplete until restoration ownership and a test route are clear. Decision aid for: Customer responsibilities remain real

AI adds specific risks

Prompt injection, poisoned documents, over-broad retrieval, unreviewed models and confident inaccurate outputs require controls beyond a conventional file server.

Source citations, permission-aware retrieval, evaluation sets, user training and human review are part of the operating design.

Exit path

Ownership should preserve a practical route out

Export Data, settings and useful records.
Rebuild Documented components and versions.
Replace Alternative runtime or supplier route.
Close Deletion, access removal and asset handling.
Portability still needs testing. Open components do not remove migration work. Decision aid for: AI adds specific risks

No certification claim without certification

The proposed appliance is not marketed as automatically UK GDPR compliant, Cyber Essentials certified, NHS DSPT compliant, SRA approved or FCA approved.

Those frameworks and sector obligations can guide the design, but organisational assurance requires its own evidence.

Questions answered

Straight answers to common questions

Is the server secure out of the box?

It can be supplied with an agreed baseline, but security depends on the final network, users, data, updates and operating ownership.

Can it be air-gapped?

An offline or air-gapped design can be scoped, including controlled update and transfer procedures. It is not a default hardware feature.

Do you provide 24/7 monitoring?

Not in the initial supply-and-onboard proposition. Any managed monitoring or response time requires a separate service agreement.

Primary-source register

Check the live rule or price before relying on it.

Reviewed 26 July 2026. These links support the dated statements on this page; they do not replace legal, tax, security or professional advice.

Put the claim to work

Turn this guidance into a testable requirement.

The brief asks about workload and operating conditions - not just budget.