← Back

AI inside a business needs boundaries.

Nexel designs each environment around explicit control of data, access, model use, workflow authority and human approval.

Detailed controls are agreed for each deployment.

01

Separate client environments

Each client environment is designed as an isolated deployment rather than a shared application database.

02

Permission before retrieval

Access rules should be enforced before relevant content is presented to the model.

03

Model control

Model choice should be explicit rather than silently changed underneath business workflows.

04

Human authority

Workflows define where approval is required.

05

Evidence

Material outputs should retain links to the information used to produce them.

06

Evaluation

A workflow should demonstrate performance against real prior work before dependence on it increases.

What this page will not do

Specific controls — encryption, identity integration, retention periods, log formats and deployment topology — are stated for your environment in the deployment documents, where they are specific enough for your IT provider to check and binding enough to be worth checking. A matrix here would be less precise and commit us to less. Send us your vendor security assessment and we will answer it directly.

Before you ask

Who you contract with, and whose law applies.

Nexel Ventures Pte Ltd is a Singapore-registered company, UEN 202333215Z, and it is the entity you contract with wherever you are. We are not going to pretend that is an Australian company when it is not.

Governing law follows you rather than us. Australian engagements are governed by Australian law and Singapore engagements by Singapore law, so an Australian firm is not agreeing to resolve a dispute in a foreign forum in order to buy software.

Data residency is fixed to the jurisdiction named in your contract. An Australian firm’s corpus stays in Australia, including the inference that reads it. If your risk committee needs an arrangement other than this one, raise it before the audit rather than after: it is a scoping question, not an obstacle.

Trust should be designed into the workflow.

Discuss your deployment