Draft The roles below are what the service uses today; the names wait in [brackets]. [Decide: customers must be told each subprocessor's name, and the site's rule keeps provider names off its pages. Either make this page the exception, or send the list with the data processing agreement.]
Subprocessors
These providers help us run the service, and handle customer data while they do. Each is bound by terms at least as protective as our data processing agreement [confirm one is signed with each], and we tell customers 30 days before adding or replacing one.
| Provider and role | Customer data it handles | Where | Safeguard |
|---|---|---|---|
| [Name] Model routing: passes each review prompt to the model that answers it | Review prompts: diffs and the code around them, house rules, dismissal reasons, scanner findings, the text of linked issues | [ ] | [ ] |
| [Names] Model providers: run the models that answer review prompts, behind the routing service or directly | Review prompts, as above | [ ] | [ ] |
| [Name] Server hosting for the review workers and the event router | Everything the security and data page lists as kept | [ ] | [ ] |
| [Name, if used] Network tunnel that carries GitHub's events to our servers | Events passing through, with pull request titles, descriptions and comments [confirm] | [ ] | [ ] |
| Paddle Merchant of record: sells the plans through its checkout, takes the payment and collects the tax | The GitHub account ID a plan is bought for, each month's seats and overage charge, and the billing contact and payment details entered at checkout | [ ] | [ ] |
GitHub isn't on this list: your code lives there already. Plans sell through Paddle checkout first. A GitHub Marketplace listing comes later, as a second way to buy, and bills under GitHub's own terms.