Privacy policy
How Precedent handles personal data
Precedent is a decision log for teams and the AI agents that work with them. This policy explains what we collect when you visit precedent.biz or use the hosted service, why, who else is involved, and the choices you have. Last updated 19 November 2026.
Who we are
The hosted service at precedent.biz is operated by IMGN Technologies (Pty) Ltd (“we”, “us”), a private company registered in South Africa, trading as Precedent. You can reach us about anything in this policy at [email protected]. The same details appear on our legal notice.
- Registered name
- IMGN Technologies (Pty) Ltd
- Trading name
- Precedent
- Legal status
- Private company incorporated in South Africa and registered with the Companies and Intellectual Property Commission (CIPC)
- Registration number
- 2015/026469/07
- Registered office
- 20 Krynauw Street, Worcester, Western Cape, 6850, South Africa
- Director
- Gordon Byron
- Information Officer
- Gordon Byron
- [email protected]
This policy covers the public website, the web application, the MCP (Model Context Protocol) endpoint that assistants connect to, the Slack and Microsoft Teams integrations, client portals, and the emails we send. It does not cover copies of Precedent that other organisations run on their own infrastructure; see self-hosted copies.
Controller or processor
Precedent is a business tool, so our role depends on the data.
- Your account, billing, security logs and the public site. We decide why and how this data is processed, so we are the controller.
- What your workspace puts into Precedent. Decisions, rationale, evidence, comments, member lists, invitations, captured chat threads and portal invitees belong to the workspace. The organisation that owns the workspace is the controller and we process that content on its instructions, under our data processing agreement. Questions about content a workspace holds about you should go to that workspace’s administrators first; we will help them answer.
What we collect
| Category | What it includes | Where it comes from |
|---|---|---|
| Account | Email address, display name, and the profile picture your identity provider shares. Which sign-in methods you have used, when you accepted the terms of service and which version, and a 30-day session cookie once you are signed in. | You, or Google or your company’s single sign-on provider when you sign in with them |
| Workspace content | Decisions and every revision of them: decision text, including how-to steps, rationale, alternatives, assumptions, review dates and conditions, evidence links, uploaded files, Area tags, Types, and the history of who did what to each record. Proposed changes to decided decisions are kept with the decision, including after they are approved, withdrawn or go out of date. | Members of the workspace, and assistants acting for them |
| Membership and access | Workspace and group membership, roles, pending invitations by email address, how a membership came about (invitation or SSO), portal invitees by email address, and legal holds. | Workspace administrators |
| Chat integrations | For Slack and Teams: your chat user id, team or tenant id and display name, so a command in chat can be tied to your Precedent account; the thread text when someone captures it as a draft; and the channel a workspace links to an Area. | Slack or Microsoft, when a workspace connects them and you link your chat account |
| Connected tools | Credentials a workspace administrator provides for Jira, Linear, GitHub, Notion, Google Drive, a model provider or a chat app, stored encrypted. Titles fetched from those tools when a link is pasted as evidence. | Workspace administrators |
| Agent connections | The AI clients you have authorised, the workspace or Area each may act in, the permissions you approved, and for every tool call: the tool name, duration, outcome, request id and any error. Not the content of the call. | You, when you approve a connection, and the client when it calls |
| Ask AI and automation usage | For each model request: provider, model, which tools ran, token counts, duration and success. We do not store the text of your questions or the model’s answers. | Generated when you use Ask AI or when automation runs |
| Notifications | Your notification preferences, and a delivery log with the address or channel, subject, item count and outcome of each email or chat message we send. | You, and our sending systems |
| Billing | Paddle’s identifiers for your customer record and subscription, the subscription status, and monthly usage counts for plan limits. Paddle holds your card details and invoices; we never see the card number. | Paddle, our merchant of record |
| Technical | Server logs with IP address, browser and timestamps, and application logs and error details that can include your user id, the workspace id and, when an error names it, an email address. | Your browser and our hosting provider |
We do not buy data about you, run advertising trackers, or use third-party analytics cookies.
Why we use it
- To provide the service you signed up for (performance of a contract): signing you in, showing you the decisions you are allowed to see, sending the emails and chat messages you or your workspace asked for, metering plan allowances and taking payment through Paddle.
- To keep the service secure and working (legitimate interests): server and error logs, rate limiting, the audit trail of who changed what, and investigating abuse.
- To meet legal obligations: tax records held by Paddle, and responding to lawful requests.
- With your consent where the law requires it, for example optional notification channels. You can withdraw consent in Settings or by contacting us.
We do not use your content or your behaviour to build advertising profiles, and we do not sell personal data.
AI features and model providers
Several features send content to a large language model. What goes where depends on how your workspace is set up.
- Your workspace’s own model account. An administrator can connect an Anthropic, OpenAI or OpenRouter account under Settings, Integrations. When one is connected, Ask AI, conflict checks, chat capture and transcript extraction use only that account. Requests go directly to that provider under your organisation’s own agreement with them, so that provider is your organisation’s processor, not ours. We store the key encrypted and never display it again.
- The app-funded default. When a workspace has no model account of its own, Ask AI, automatic conflict checks, Slack and Teams capture, transcript extraction, Slack suggestions and workspace setup suggestions use a model account we pay for, reached through Vercel AI Gateway. The request contains only the content needed for that task: the decision text being checked or the thread being captured. An Ask AI request contains the question and the earlier turns of the conversation, the name and role of the person asking, the workspace name, and the records Ask AI reads to answer. Those records are decisions with their evidence excerpts, Areas and review items, and they can include the names or email addresses of decision owners, approvers and Area members, and of people quoted in a captured thread.
- Search embeddings. Decision text and search queries are turned into vectors so that search can find decisions by meaning. On the hosted service this uses an OpenAI embedding model reached through Vercel AI Gateway, for every workspace, whether or not it has its own model account. Self-hosted copies can run embeddings inside their own deployment instead.
We do not use your content to train models, and the API terms of the providers we use do not allow them to train on it either. Model providers process each request to return a response. Their own retention and security terms apply and are available from us on request. Model output is a suggestion: a person confirms every decision.
Where data is stored
The hosted service at precedent.biz keeps its database and uploaded files in Ireland, in the European Union, and runs the application there. Precedent can run as separate deployments per region, each with its own database and file store. A workspace is stamped with the region it was created in and is served only there; its content does not move between regions.
Several of our providers are based in or operate from the United States; the sub-processor list gives each one’s regions. Where personal data leaves the European Economic Area, the United Kingdom or Switzerland, we rely on the European Commission’s Standard Contractual Clauses and the UK International Data Transfer Addendum, or on an adequacy decision such as the EU-US Data Privacy Framework where the provider is certified. Copies of the relevant terms are available on request.
How long we keep it
- Your account exists until you ask us to delete it. Sign-in links expire after 15 minutes and sessions after 30 days.
- Workspace content is kept for as long as the workspace exists, or as its administrators decide. Administrators can set a retention policy that permanently deletes archived decisions after a chosen period and trims operational logs; decided decisions are never purged automatically. A legal hold pauses every purge. Administrators can also export the whole workspace and delete it entirely under Settings, General. The export is a JSON file that links to uploaded files rather than including them, so download any files you need before deleting the workspace.
- Uploaded files are deleted from storage when the evidence, decision or workspace they belong to is deleted.
- Deleted data lingers in encrypted backups for a limited period before those backups roll over.
- Server and error logs are kept for 30 days by Vercel, our hosting provider.
- Billing records are kept by Paddle for as long as tax law requires.
Security
Data is encrypted in transit and at rest. Model account keys and integration credentials are encrypted again at the application level before they reach the database. Every list, search, export and agent tool call applies the same visibility rules, enforced in the database with row-level security as well as in the application. Assistants connect through OAuth 2.1 with PKCE and act with the permissions of the person who approved them, who can revoke that approval at any time. Audit exports are signed so they can be verified independently.
No system is perfectly secure. If we learn of a breach affecting your personal data we will notify affected workspace administrators and, where required, the relevant authority, without undue delay. To report a vulnerability, write to [email protected].
Cookies
We set only the cookies needed to run the service: a session cookie that keeps you signed in for up to 30 days, and short-lived cookies that protect sign-in and OAuth flows against forgery. They are strictly necessary, so no consent banner is shown. We do not set advertising or analytics cookies, and we do not use local storage to track you across sites. The cookie policy lists every cookie and storage key by name, with its purpose, provider and lifetime.
Your rights
Depending on where you live, you may have the right to access the personal data we hold about you, to correct it, to have it deleted, to restrict or object to how we use it, to receive a copy in a portable format, and to withdraw consent where we rely on it. You also have the right to complain to a data protection authority, in the EU typically the one in the country where you live or work, and in the UK the Information Commissioner’s Office.
Where a workspace is the controller of content about you, we will pass your request to its administrators and help them respond, unless you ask us not to.
Children
Precedent is for workplaces. It is not directed at anyone under 16, and we do not knowingly collect personal data from children. If you believe a child has created an account, contact us and we will delete it.
Self-hosted copies
Precedent’s source can be run by other organisations on their own infrastructure. We have no access to those deployments and this policy does not apply to them. The organisation running the copy is responsible for its own privacy notice.
Changes to this policy
We will update this page when our practices change and revise the date at the top. For material changes, such as a new category of data or a new purpose, we will tell workspace administrators by email or in the product before the change takes effect. Earlier versions are available on request.