Skip to content

Data, privacy and AI

The short version: all data stays on Atlassian infrastructure. The vendor never sees it.

All app data is stored in Forge hosted storage on Atlassian infrastructure, isolated per site. There are no third-party services, no analytics, no tracking and no external data flows of any kind. Forge hosted storage follows the data residency settings of the host product.

Per reminder: the owner’s account ID, the ticket key or JQL text, the trigger definition (including the AI condition text if used), notification channels, the owner’s profile language, status and timestamps. Additionally: in-app notifications (subject and timestamp), an audit log per reminder, and anonymous usage counters for the AI quota. Ticket contents are not copied into app storage.

Nothing stored by the app: Forge hosted storage is operated by Atlassian and is not readable by the vendor. Operational platform logs are deliberately kept free of personal data and content; they contain technical identifiers and error messages, but no comment texts, no AI reasoning and no full account IDs.

Reminders with an AI condition evaluate new comments via the Forge LLMs API: models hosted and operated by Atlassian on Atlassian infrastructure. Comment texts do not leave the Atlassian platform. Atlassian’s Acceptable Use Policy and AI moderation apply. There are no AI costs for the customer organization; model usage is billed to the app vendor.

Every AI decision is written to the reminder’s audit log in the site’s own app storage, where users can review it.

  • Monthly pool per site: licensed users × 20 (Standard) or × 200 (Advanced) evaluations, with a guaranteed minimum pool for small sites.
  • All users share the pool freely, following the same model as Atlassian Automation. Usage is visible to everyone in the app.
  • When the pool is exhausted, AI reminders pause until the first of the next month. Affected users are notified once; all other functionality continues normally.
  • Cost controls built in: comments are pre-filtered by JQL before any AI call, and trivial comments (“ok”, “+1”) are never evaluated.

Two scheduled runs keep the shared AI quota from being wasted. Neither of them deletes anything.

When a Jira account is deactivated, a daily run pauses that person’s reminders. Watchers of accounts that have left would otherwise keep running and keep drawing from the site’s shared AI quota, and for AI triggers the quota is consumed before the model is even called. The app only pauses on an explicitly deactivated account; if the state cannot be determined, the reminder is left alone. Reactivating the account deliberately does not reactivate the reminders: after a longer absence old watchers are often no longer wanted and would immediately start consuming quota again. Each affected entry carries a Reactivate button, so the person decides which ones they still need.

Expiry dates are checked hourly. Users can give repeating reminders and filter alerts an expiry date, and no expiry is preselected. Expired rules are paused and can be extended with one click.

Users can delete individual reminders anytime and remove all of their data with one action inside the app. After uninstalling, Atlassian deletes the app’s storage according to the Forge platform retention period (currently 28 days). The vendor holds no copies.