AI Duplicate Detector & Issue Assistant - Privacy Policy
Updated: 2026-07-16
This Privacy Policy explains how Pokutsoft handles data in connection with AI Duplicate Detector & Issue Assistant for Jira (the "App"), a Forge app for Atlassian Jira. This policy covers this App specifically; Pokutsoft's other Atlassian Marketplace apps are covered by their own, substantially similar, privacy policies published alongside each app.
1. Who we are
Pokutsoft develops Forge-based Marketplace apps for Atlassian products (Jira, Confluence, Jira Service Management). Pokutsoft operates as the vendor of record for this App on the Atlassian Marketplace.
2. What data we process
The App runs under Atlassian's "Runs on Atlassian" model: data does not leave the Atlassian cloud (there is no external egress). All reads and writes are performed on behalf of the current user, so the App only ever accesses or changes issues that user already has permission to work with. Specifically:
- Processed (transient): The current Jira issue's summary, description (truncated), project key and, for some AI actions, its recent comments (up to 10, each truncated) — read from the Atlassian API at the moment you open the panel or run an action.
- Processed (transient): A JQL search over other open issues in the same project, to gather a small set of candidates (titles truncated) that the AI model then compares. The project-page backlog scan reads titles of a limited set of open issues (up to 40) in a single pass.
- Written to Jira (only when you act): When you explicitly choose to, the App writes to Jira on your behalf — adding a comment, creating sub-tasks, updating the description, setting story points, linking issues, moving watchers, or transitioning a duplicate to closed. When a new issue is created, the App may automatically add a comment listing existing duplicates it found.
- Stored (limited retention): The App caches its generated duplicate result — candidate issue keys, titles, match scores, and short reasons — in app-namespaced Forge storage for up to 24 hours, keyed to the issue and a hash of its content, so the same, unchanged issue is not re-analyzed unnecessarily. AI-action results are not stored; they are returned to you and discarded.
- Not stored: The App does not permanently store Atlassian account personal data (email address, username, account ID, avatar).
3. AI / LLM usage
The App uses Atlassian's hosted Forge LLMs (Anthropic Claude) API. Depending on the feature, the model is asked to judge whether candidate issues are duplicates, group likely duplicates in a backlog, or produce the output of an AI action you run (for example a summary, acceptance criteria, an estimate, or a translation). The prompt sent to the model is processed within Atlassian's infrastructure; no data is sent to any external LLM provider. Your content is not used to train any model.
4. Data retention and deletion
Cached results are automatically deleted after 24 hours, or invalidated immediately if the underlying issue content changes. When the App is uninstalled, Atlassian removes the App's namespaced storage.
5. Data subject rights (GDPR/KVKK)
Because the App does not store Atlassian account personal data, deletion and correction requests are generally handled through Atlassian directly. For any App-specific request, contact us at the address below.
6. Sub-processors
- Atlassian (hosting, the Forge platform, and Forge LLMs). Data remains within Atlassian's infrastructure boundary.
7. Changes to this policy
This policy may be updated from time to time. The current version is always published at this page.
8. Contact
support@pokutsoft.com · Vendor: Pokutsoft