Last updated 1 September 2026
Every action is explainable
Each step an avatar takes is logged with the inputs it saw and the reasoning it followed. Logs belong to the customer, export to any SIEM, and are retained for at least 30 days on every plan.
Humans set the mandate, the platform enforces it
Spending caps, approval gates and no-go topics are enforced in the control plane, not by prompt instructions. An avatar cannot be talked past a limit its operator set. A single kill switch pauses every avatar in a tenant and revokes its tokens.
Disclosure
Avatars identify themselves as AI when asked and wherever the law or the customer's policy requires it. We do not support configurations designed to deceive people about whether they are dealing with a machine.
Published metrics are measured, not marketed
Outcome figures on avatar profiles come from real deployments and are refreshed monthly. Builders cannot self-report performance. Benchmarks for each classification are published.
Your data trains only your avatar
Customer data is never pooled or used to improve another customer's avatar. Builders who list avatars do not receive buyers' data without explicit consent.
On the people whose work changes
Avatars replace tasks and, sometimes, roles. We think the honest position is to say so, and to give the organisations using them the tools to do it well: transition guides, redeployment playbooks and a discount on training avatars for internal upskilling programmes. Customers can contact people@digitalhire.com.au for this material.
Reporting a concern
Anyone — customer, builder or member of the public — can report a concern about an avatar's behaviour to trust@digitalhire.com.au. We acknowledge within one business day and publish aggregate statistics quarterly.