Questions CTOs and CISOs ask first.

It depends on where Polis runs. In your environment, it stays there. With a dedicated instance we operate in the EU, it is processed on that instance, which serves only you.

Either way, model calls go to your own provider, only for the data classes you allow, and Polis stores no prompt or response content.

The AI Developer uses Claude via the Anthropic API or Vertex AI. Other roles can use Claude, Gemini (Vertex AI or the Gemini API), Amazon Bedrock, Azure OpenAI or OpenAI. All on your own contract.

Merge to the default branch, deploy, or change Polis’s own settings. That is enforced in code, and no setting changes it.

Safety rules such as “never grant admin rights” are on by default. Only your Owner can switch one off, with a reason, and a banner stays on every page while it is off.

Polis does not. What your model provider may do is governed by your contract with them.

Work arrives as draft pull requests, so nothing reaches the default branch without your review. Every decision, approval, tool call and model call is in the audit log, and one switch pauses all AI work.

In your environment, on-premises or in your cloud, on a server or VM you provide.

Or on a dedicated instance we operate in the EU: one customer per server, never shared.

A person. The AI proposes lessons, and a person approves each one before it is used. You can edit, retire or delete any lesson.

Polis has no security certification yet (such as ISO 27001 or SOC 2). The security brief describes each control.

Read the security brief

AI employees are system accounts with roles and limits. Polis records who approved what, for the audit; it doesn’t measure or rate staff. We’re happy to bring your works council into the security deep-dive.

Talk to us. Model usage is billed by your provider, to you.

No. It sets no cookies, runs no analytics and loads nothing from third parties.

Privacy