Who holds your data
Arbiter does. This is a product demonstration that Arbiter runs for itself: no company invited you, no application exists, and nothing you do here reaches a hiring team. Questions and requests about your data go to support@getarbiter.io.
What is collected
- Your business email address, which you gave to request the link. It is kept as a sales lead and as the record that this address has had its one demo. The industry and the exercise you chose are kept beside it.
- The code and files you write during the session, with automatic snapshots as you go.
- The commands you run in the terminal and whether they succeeded, the files you save, and when this browser tab loses and regains focus.
- Your prompts to Pair and its replies, if you use it — your words exactly as typed, including anything you paste into Pair. Pair sees no files and no brief. Your prompt text is sent to Anthropic, the model provider.
- The size of anything you paste into the editor or the terminal — its length and line count, never the text itself — and each time the paste limit refuses a paste.
- The answers you give at submit, if you give them: how you rate your own solution, and how much of the work you say was AI's.
- A record that you were shown this policy and the demo notice, and which version.
No CV is read, and none is asked for. No name is asked for.
What is not collected
No screen recording, no camera, no microphone. No biometric data of any kind, and no keystroke logging: nothing you type is recorded beyond the code, the commands and the prompts listed above, and typing rhythm or cadence is never analysed. Nothing about other applications or other browser tabs, beyond noting when this tab loses focus. A paste is measured, not kept — what you paste into a file becomes part of that file, and files are kept with the session; what you paste into Pair becomes part of that prompt, and prompts are kept with the session.
Automated processing, and no decision
Software runs your code against a test suite and calculates the score on your report. The report also describes how you worked, and a model may read the code you wrote to draft the interview questions it shows — the way it would for a real candidate, because that is the product being demonstrated. No model scores your work. No person sees your report unless you send them the link. No decision about you — hiring or otherwise — is made by anyone, so you are not subject to a decision based on automated processing.
How long it is kept
Your code, the session record, your Pair transcript and your report are deleted between 24 and 30 hours after the link was sent to you: the sweep runs every six hours against a 24-hour age. The report link stops working at the same moment. What remains afterwards is an anonymised skeleton — that a demo was run, on which exercise, scored how, taking how long — with nothing in it that is or can be connected to you.
Two things stay. Your business email, with the industry and exercise you chose, stays on our lead list until you ask us to remove it. The record that you were shown these notices stays as audit material; once the session is swept it carries no identifier that points back to you.
Who else touches the data
E2B runs the sandbox your code executes in. Anthropic receives your Pair prompts and, after grading, may receive your code to draft the interview questions. Resend delivers the email with your link. None of them receives your data for any other purpose.
Your rights
Depending on where you live, you may have the right to access your data, correct it, delete it, object to its processing, or receive a copy. For the session, the simplest way is to wait: it is gone within 30 hours. For your email, write to support@getarbiter.io and we will remove it from the lead list.