Security decisions, instance by instance.

The target deployment isolates each client instance. Its database, storage, access routes and authorised people are documented and verified before productive processing.

Data regions, subprocessors and any third-country transfers are confirmed in writing for the instance before productive processing. The current processing chain belongs in its security review.

The model provider, processing terms and data route are specified for each instance before productive use. No provider or training condition is assumed from the Core architecture alone.

In bounded Core tests, an eligible human approves the exact stored output before deterministic delivery. An enabled client workflow must pass its own deployment checks.

The tested Core records execution basis, check results and approvals. The scope and method of audit export are confirmed for each deployment.

Governance features such as versioned evidence, checks and human approval can support internal control processes. This is not a compliance badge. Applicable obligations depend on the concrete use.

Brand Truth sits in the Brand Definition, a documented, machine-readable format. Export, retention, deletion and surviving audit evidence are specified in the client contract and verified for the deployment.