Home / Approved AI

Say yes to AI on purpose: ask, check, decide.

Anyone can ask for a tool, Sargha runs the checks it can, and a different person decides with those checks in front of them.

The Sargha Approved AI catalog: counts of approved, approved with conditions, not allowed and waiting, then a table of tools with the decision, terms, owner, who decided and the review date — The catalog your people see: each tool with its decision, its terms and a review date. The checks inform. They never decide. Sample data from the Sargha sandbox.

One list of what you decided

Approved, approved with conditions (for example internal data only), not allowed, or waiting. Each entry has an owner and a review-by date.

Checks that help, not decide

Sargha runs automatic checks and shows the results. A clear problem is flagged, and a person still records the decision.

Nobody approves their own request

A different person decides. Requests can be sent back for more information.

Never just no

An entry can name approved alternatives, so the answer points to something people can use.

Shows up where the question is asked

A finding for an unapproved tool says what to use instead. Approving or denying updates the lists sensors use: warn by default, block where you have switched enforcement on. Each entry has a review-by date, and a review queue returns it to its owner when the date passes.

Questions

Common questions.

Who can ask for a tool?

Anyone with a console account can ask for a tool, a tool server, a skill, a model, an extension or a service.

Can the checks block something on their own?

No. A clear problem is flagged, but a person still records the decision.

What does an entry look like?

Each entry shows the decision (approved, approved with conditions, not allowed or waiting), the terms, the owner, the decision date and the review-by date, so anyone can see what was decided and when it is due for review.

What happens after a decision?

Approving or denying updates the approved and blocked lists that sensors use. What happens on a machine then depends on its mode: warn by default, block where you have switched enforcement on.

Can a request be sent back?

Yes. A person can send a request back for more information instead of deciding straight away.

Where do people see the answer?

In the finding, component or tool-server panel for an unapproved tool, which names the approved alternatives, for example use instead of personal ChatGPT accounts. On a phone with AI apps in personal or off-network use, the panel points to the catalog and to your device-management tool.

What you need

A console account for anyone who will ask, and an administrator to decide.

For your technical team

Policy, enforcement modes and data-handling detail are in the technical pack, shared with your team under NDA.

Request the demo and technical pack

Get started

See the AI on your fleet in about a week.

Run the sensor across your estate and receive an AI risk report and an AI Bill of Materials. The assessment is free and carries no commitment.

Request a free assessment → Book a 30-minute call