Skip to content

Sub-processors

Sub-processors

IMGN Technologies (Pty) Ltd operates Precedent (see the legal notice). To run the hosted service at precedent.biz we use the providers on this page. Each one processes personal data for us, including content that workspaces store in Precedent, and only to deliver its part of the service. This page accompanies the privacy policy. Last updated 12 November 2026.

Current sub-processors

Every provider below is in use on precedent.biz today. The terms link goes to the provider’s own data processing terms.

ProviderPurposeData processedRegionsTerms
VercelRuns the application, serves the website and runs scheduled jobs.Everything the application handles passes through it while a request is served, including account details and workspace content. Request logs with IP address, browser and the page requested, and the application’s own logs, which can include error messages, user ids, workspace ids and, when an error names one, an email address. Both are kept for 30 days.Application functions in Dublin, Ireland. Requests enter at the nearest point of Vercel’s global network.Vercel DPA
SupabaseManaged PostgreSQL database, and S3-compatible storage for uploaded evidence files.All stored account and workspace data: decisions and their history, members and invitations, encrypted integration credentials, audit records and uploaded files.Ireland (AWS eu-west-1).Supabase DPA
CloudflareDNS, TLS and a proxy in front of precedent.biz that protects the site from abuse.All traffic to and from precedent.biz while in transit, including IP addresses and request content.Global network. Each request is handled at the Cloudflare data centre nearest the visitor.Cloudflare DPA
ResendSends transactional email: sign-in links and the notification emails the product sends, such as review reminders and the weekly digest.Recipient email address and name, and the message, which can include workspace names and decision titles.Sends from Ireland (AWS eu-west-1).Resend DPA
Vercel AI GatewayRoutes app-funded AI requests to a model provider: search embeddings for every workspace, and, for workspaces without their own model account, Ask AI, automatic conflict checks, chat and transcript capture, Slack suggestions and workspace setup suggestions.The text each task needs: decision titles, summaries, rationale, alternatives and assumptions; the chat thread or transcript being captured; Ask AI questions with the name and role of the person asking, and the records read to answer them, which can include the names or email addresses of workspace members and of people quoted in evidence; search queries.Not pinned to a region.Vercel DPA
OpenAIModel provider behind Vercel AI Gateway for the two models in use: GPT-5.6 Luna for Ask AI and automation, and text-embedding-3-large for search.The content Vercel AI Gateway forwards for each request.Not pinned to a region.OpenAI DPA
Microsoft AzureModel provider behind Vercel AI Gateway, which may serve either model in use through Azure OpenAI.The content Vercel AI Gateway forwards for each request.Not pinned to a region.Microsoft DPA
Amazon Web ServicesModel provider behind Vercel AI Gateway, which may serve GPT-5.6 Luna through Amazon Bedrock. Not used for embeddings.The content Vercel AI Gateway forwards for each request.Not pinned to a region.AWS DPA
Google WorkspaceEmail for [email protected], where support, privacy and security requests arrive.Your email address and whatever you include in a message to us.Not pinned to a region.Google Workspace DPA

Vercel AI Gateway chooses which of the three model providers serves each request. Requests are not yet restricted to a particular provider, a region, or providers that keep no copy of the request. We are deciding those requirements, and when they are enforced this table and the change log will say so.

This list covers the hosted service only. Organisations that run their own copy of Precedent choose their own providers.

Where data is stored

The database and uploaded files for precedent.biz are stored in Ireland, and the application runs there. Traffic passes through Cloudflare’s and Vercel’s global networks on its way to and from Ireland, and AI requests may be processed wherever the model provider serves them. Several of these providers are based in the United States. The privacy policy explains the safeguards we rely on when personal data leaves the country it was collected in.

Not our sub-processors

Some services handle workspace data because the workspace chose to connect them. They act for the customer, under the customer’s own agreement with them, or as independent third parties. They are not sub-processors that we engage, and they are not listed above.

  • Model accounts a workspace brings. An administrator can connect the workspace’s own Anthropic, OpenAI or OpenRouter account under Settings, Integrations. Requests that use it go directly to that provider under the workspace’s agreement with them. OpenAI appears in the table above only for requests we pay for through Vercel AI Gateway.
  • Chat tools. Slack and Microsoft Teams, when a workspace installs its own Slack app or adds a Teams outgoing webhook to capture threads and post notices.
  • Work tools. Jira, Linear, GitHub, Notion and Google Drive, when a workspace connects them or links them as evidence.
  • AI assistants and agents. The clients a member authorises to connect to Precedent over MCP.
  • Sign in with Google. When you choose it, Google acts as your identity provider under its own terms.

Supported but not in use

The Precedent software can also take payment through Paddle and offer single sign-on through WorkOS. Neither is configured on precedent.biz today, so neither receives data from the hosted service. If we turn one on, it will be added to the table, with an entry in the change log, before it receives personal data.

Change log

  • : Removed Sentry. precedent.biz no longer sends error reports, logs or request timings to it; errors and logs stay with Vercel, which hosts the service.
  • : No sub-processor added or removed. Sentry now also receives how long each request to precedent.biz takes, and how long each call it makes to another service, such as Slack or an AI provider, takes, with the same user, workspace and role ids as error reports and without queries or content.
  • : Sentry now receives error reports and logs from precedent.biz, earlier than the 14 December 2026 start announced the same day.
  • : Announced Sentry, for error monitoring. It will start receiving error reports and logs from precedent.biz on 14 December 2026, and stores them in Frankfurt, Germany. Until then it receives no data from the hosted service. To object, write to [email protected] before that date.
  • : No sub-processor added or removed. Vercel AI Gateway and the model providers behind it now also process Ask AI questions, and the records read to answer them, for workspaces that have not connected their own model account.
  • : Published the list: Vercel, Supabase, Cloudflare, Resend, Vercel AI Gateway and the model providers behind it (OpenAI, Microsoft Azure and Amazon Web Services), and Google Workspace.

Getting notified

Follow the Atom feed to hear about changes. Every change log entry appears in it on the day it is published. Any feed reader can subscribe, and so can services that forward new feed entries by email.

We publish a change here before a new sub-processor starts processing customer personal data. The data processing agreement sets the notice period and your right to object.

To object to a new sub-processor, or to ask about any provider on this page, write to [email protected].

Discard unsaved changes?

Your changes have not been saved. Stay to keep editing, or discard them and leave.