Secure AI Readiness Check | Hexdata Technologies

Purpose & Information

Confirm what the system is intended to do and what organizational information may reach it.

Control objective A bounded use case with known data exposure.

Is the use case defined, owned and connected to a measurable outcome? Yes Partly No

Is sensitive, confidential or regulated data identified? Yes Partly No

Have model retention, training, residency and confidentiality terms been reviewed? Yes Partly No

Access Boundaries

Determine whether people, services and retrieved content remain inside intended authorization boundaries.

Control objective Traceable, least-privilege access.

Are user, service and model permissions connected to defined identities? Yes Partly No

Does retrieval preserve source-system authorization before content reaches the model? Yes Partly No or not confirmed

Safe Behaviour

Establish how unsafe behaviour, unintended disclosure and consequential outputs will be detected and controlled.

Control objective Tested safeguards with accountable human oversight.

Has the workflow been tested for manipulation, disclosure and unsafe tool actions? Yes Partly No

Are human approval and escalation requirements defined? Yes Partly No

Can relevant access, model activity, outputs and administrative changes be traced? Yes Partly No

Production Control

Confirm that performance, cost, support, continuity and change responsibilities are understood before scale.

Control objective An operable platform with visible ownership and cost.

Are performance, usage growth and model or infrastructure costs understood? Yes Partly No

Are support, incident, change, continuity and retirement responsibilities assigned? Yes Partly No

Before AI reaches production, test the foundations around it.