Responsible AI at Vionx

Effective date: 2 November 2026

Vionx OS lets companies hand repetitive work to AI agents. These are the principles built into the product, and what we ask of customers.

People stay in charge

  • Agents act only with the tools and permissions a workspace admin gives them.
  • Actions with an external effect, such as sending an email, posting, paying or updating a record, can be set to wait for a person's approval.
  • Every run is logged with its steps, inputs, outputs and cost, so a person can see what an agent did and why.

Your data is not used for training

We do not use customer data, or data from connected tools, to train or improve generalised AI models. Where a model provider offers a no-training option, we use it.

You choose the model

Each workspace chooses which model runs each step: a cloud model, or a local model on the organisation's own hardware where data does not leave the building. Admins can limit the workspace to approved models. Current providers are listed on the Subprocessors page.

Know the limits

  • AI output can be wrong, incomplete or out of date. Review important output before relying on it.
  • Agents that answer customers can be limited to your own documents and told to hand over to a person when unsure.
  • Do not use agents to make decisions with legal or similarly significant effects on people without human review (see the Acceptable Use Policy).

Transparency

We recommend telling customers when they are talking to an AI agent. Customer-facing agents in Vionx OS can introduce themselves as automated assistants.

Arabic first

We test agents in Arabic as well as English, and check that answers are accurate and appropriate for the Saudi context.

Report a concern

Email hello@vionx.ai with the subject “Responsible AI”.