Jira data can be exported and licensed for AI training only within three layers of control: the company's own rights in the issues, Atlassian's terms and settings, and the rights of the people named in them. Atlassian's data contribution for AI took effect on August 17, 2026, and eleven buyer and intermediary programs name Jira among the systems they want (checked November 9, 2026).
Jira is one source system in AI training data licensing, and the answer to "can the company license it" depends on the plan, the export route and what the issue text contains. A Jira Cloud backup can be created once every 48 hours with attachments, and Atlassian restores only from backups no older than 30 days.
The DataPacts Editorial Team states each plan setting, term date and program ask below with its check date; the method is explained in how we source, date and check every figure.
What Jira Issue History Holds and Why AI Buyers Ask for It
Jira issue history records how a team plans, discusses and fixes work over years, and AI buyers ask for it mostly as context next to code and other company records.
Brendan Foody, who describes himself as Mercor's CEO, states in a LinkedIn post that Mercor has partnered with more than 300 enterprises across 90 systems, with Jira at 35 % of them (Slack 55 %, GitHub 53 %; checked November 9, 2026). That is his claim, not an audited count. DataVendor, a marketplace, states that it wants issue or ticket history from Linear and Jira as context sent with a codebase (checked November 9, 2026), and Mercor's game-studio listing asks for chat, tickets, code and shared drives (checked November 9, 2026).
Turing prices tickets only when they come with a codebase (see below), so the value of Jira history on its own is unsettled. The wider question of issue trackers as a data family is covered in selling issue trackers and project histories for AI training; this page covers what is specific to Jira.
Who Controls Jira Data in a License
Three parties decide what Jira data a company can license: the company as customer, Atlassian as the platform, and the employees, customers and contractors whose names and words fill the issues.
Atlassian Corp (SEC CIK 0001650372) is the platform. The ownership and AI-use wording of the customer agreement is not covered on this page and should be read from the agreement itself.
Table 1 compares the four layers of control over Jira data (as of 2026-11-09).
| Layer | Who | What the record shows | Effect on a license | Source |
|---|---|---|---|---|
| Customer | the licensing company | its customer agreement with Atlassian (wording not covered here) | sets what the company may do with exports | Atlassian customer terms |
| Platform settings | Atlassian Corp | AI data contribution since 2026-08-17; opt-outs by plan | metadata may already be used by Atlassian on Free, Standard and Premium | atlassian.com/trust/ai/data-contribution |
| Platform apps | Marketplace app developers | may not sell End User Data, s. 9.2, effective 2025-12-01 | an app developer is not a resale route | developer.atlassian.com |
| People in the issues | staff, customers, contractors | German employee data rule names Jira archives | consent, notice or removal before delivery | bundesarbeitsgericht.de |
Atlassian's AI Data Contribution, Plan by Plan
Since August 17, 2026, Atlassian has drawn on customers' metadata and in-app data to improve its products and AI, and what a company can switch off depends on its plan.
Table 2 compares the opt-outs and defaults by Jira plan, as stated on Atlassian's data contribution page (as of 2026-11-09, in effect since 2026-08-17).
| Plan | Metadata contribution opt-out | In-app data contribution default | In-app opt-out | Source |
|---|---|---|---|---|
| Free | not possible | on | possible | Atlassian trust page |
| Standard | not possible | on | possible | Atlassian trust page |
| Premium | not possible | off | possible | Atlassian trust page |
| Enterprise | possible | off | possible | Atlassian trust page |
Buyers write terms that assume the seller has granted nothing to anyone else. micro1's Data License Master Terms (updated November 8, 2026) contain a no-prior-grant warranty in s.5.1, and Labelbox states that Prism licenses are exclusive by default; exclusive vs non-exclusive data licenses explains how such terms treat earlier grants. Whether Atlassian's metadata contribution conflicts with such a warranty is not settled, so it is a question to put to the buyer and to counsel, not a conclusion.
Before any license, the company records its plan, its contribution settings and the date each setting was last changed.
Marketplace Apps and the Ban on Selling End User Data
Atlassian's Developer Terms bar Marketplace app developers from selling any End User Data (s. 9.2, effective December 1, 2025), so an app installed in Jira is not a lawful route for its developer to resell the data it reads.
The section binds developers. A company licensing its own Jira history is not the app developer, and its license runs under its own customer terms. Third-party app data is also excluded from the Jira Cloud backup, a point that matters for buyer connectors vs native exports.
How Jira Data Leaves the Instance
Jira data reaches a buyer by one of three routes: a backup or export the company makes itself, a read-only connection the buyer or intermediary runs, or a snapshot prepared inside the company's environment.
Table 3 compares the routes (as of 2026-11-09).
| Route | Who runs it | What the record shows | Source |
|---|---|---|---|
| Jira Cloud backup | the company | every 48 hours with attachments; restore from backups up to 30 days old; excludes automation flows, third-party app data, Assets, among others | support.atlassian.com |
| Issue search export or REST API | the company | limits not stated here | n/a |
| Read-only connection | program (example Centific) | in-scope systems connected read-only | centific.com, checked 2026-11-09 |
| Snapshot prepared in the company's environment | program (examples Scale AI, Collinear AI) | anonymized before extraction (Scale AI); one-time snapshot (Collinear AI) | scale.com, collinear.ai, checked 2026-11-09 |
Cloud Backup: Interval, Restore Window and Exclusions
A Jira Cloud backup with attachments can be created once every 48 hours, and Atlassian restores only from backups no older than 30 days (since January 22, 2026).
Among the items the backup leaves out are these three (checked November 9, 2026):
- automation flows
- third-party app data
- Assets
A buyer that asks for "everything in Jira" receives what the backup holds, so the data manifest should list the exclusions.
Other Export Routes and What Is Not Recorded
Issue search exports and the Jira REST API are the other routes, and this page does not state their limits because they are not recorded from Atlassian's documentation. CSV and Excel exports from issue search exist, and spreadsheet tutorials are outside this page. Jira Data Center and Server mechanics are likewise not covered.
Read-Only Connections and Snapshots Offered by Buyers
Several programs offer to read Jira in place rather than receive a raw export, and their published process decides who sees identifiable data first.
Scale AI states that it anonymizes inside the company's environment before extraction and uses per-package approval gates. Centific states that in-scope systems are connected read-only and that "You license your data. You never hand it over." Collinear AI states that it takes a one-time snapshot and the seller keeps ownership (all checked November 9, 2026).
Troveo, a marketplace, states that no OAuth access or install is required, and Avelence, a broker, states that it has no default system access. Both handle data on the seller's behalf and do not buy it (checked November 9, 2026).
What to Remove From Jira Issues Before Delivery
Jira issues carry personal data, customer information and secrets in free text, so removal starts with the fields buyers do not need and ends with comments and attachments. Methods for PII removal and de-identification apply to issue text as to any other text.
The removal list for a Jira export covers:
- assignee, reporter and watcher names and emails
- customer names and portal requests in service desk projects
- credentials, tokens and URLs pasted into comments
- attachments such as screenshots, logs and exports
- links to other systems
Customer-side terms for service desk portal data are not recorded here and sit in the company's own contracts.
Who pays for the work differs by program. Turing states that it covers de-identification costs, and Ooak states that it deletes raw data within 30 days of delivery (both checked November 9, 2026). In Germany, the Federal Labour Court (BAG, 8 AZR 209/21, decided May 8, 2025) disapplied § 26(1) BDSG; employee data in archives such as Jira then needs a GDPR basis, consent or a works agreement under Art. 88(2) GDPR. Employee notice and consent covers the steps before staff comments leave Jira.
What Buyers and Intermediaries Ask For in Jira Records
Eleven buyer and intermediary programs name Jira or Jira-style tickets among the records they want, and each publishes different terms, all checked November 9, 2026.
Table 4 compares them alphabetically (as of 2026-11-09); every value is the program's own statement.
| Program | Role | Jira-related records named | Terms that matter for Jira data | Source |
|---|---|---|---|---|
| Avelence seller advisory | broker (intermediary) | Jira, Linear, Asana, Notion | no default system access; de-identification on a separate copy | avelence.com |
| Centific Enterprise Data | data vendor | Confluence, Notion, Jira, Asana among 188 supported apps | read-only connection; "You license your data. You never hand it over." | centific.com |
| Collinear AI paid licensing | RL-environment builder | private repos with PRs and tickets; Jira | one-time snapshot; seller keeps ownership | collinear.ai |
| Corpus Bridge | introducer (intermediary) | GitHub, GitLab, Jira | brokerage partner approaches labs; seller pays a success fee after closing | corpusbridge.com |
| DataVendor (HUD) | marketplace (intermediary) | issue or ticket history (Linear, Jira) as context for codebases | vendors set the price; platform fee 20 % of earnings | docs.datavendor.ai |
| Labelbox Prism | data vendor | cross-system interaction data (Salesforce, Jira, Slack) | exclusive by default; trained models "cannot be unlearned" | prism.labelbox.com |
| Mercor | data buyer | Jira among tools in the software-company and game-studio listings | listing ranges USD 50,000 to 500,000 (software) and USD 100,000 to 1,000,000 (game studio) | api.data.mercor.com |
| Ooak Data partner program | RL-environment builder | Jira, Slack, Notion, SharePoint, Teams | seller chooses systems and date ranges; raw data deleted within 30 days of delivery | ooakdata.com |
| Scale Data Partnerships | human-data platform | Jira, Confluence, Asana, Notion, Linear | anonymizes inside the company's environment; per-package approval | scale.com |
| Troveo business data | marketplace (intermediary) | Jira, Linear, Asana, Notion, Figma | licenses on the seller's behalf; no OAuth access or install required | troveo.ai |
| Turing Project Lazarus | human-data platform | tickets (Jira, Linear) | 90-day exclusive selling window; USD 50 per artifact only bundled with a codebase; covers de-identification costs | lazarus.turing.com; turing.com |
Prices are the programs' claims. Mercor states ranges of USD 50,000 to 500,000 for its software-company listing, published October 17, 2026, with Jira named among the tools, and Turing states USD 50 per artifact (newsletter of March 25, 2026) only when tickets are bundled with a codebase. No seller-confirmed price for Jira records is on the public record. The full set of AI data partnership payouts and program terms compared sets these claims beside other programs.
Avelence, Corpus Bridge, DataVendor and Troveo place or list data for sellers and are not the licensee of last resort. The table lists programs by name and implies no ranking or endorsement.
Licensing Jira History: Benefits, Risks, Downsides and Alternatives
Licensing Jira history can earn money from records a company already keeps, and it can also expose staff, customers and secrets, so not licensing remains a valid result.
Benefits
- Income from records the company already keeps, as the programs' listed ranges and per-artifact prices claim.
- Some programs carry preparation costs, as Turing states for de-identification.
- A read-only route keeps the raw data in place, as Centific states.
Risks
- Personal data and employee rules, such as the German rule above.
- Secrets pasted into comments.
- Atlassian contribution settings already in effect on lower plans.
- A trained model cannot unlearn the data, as Labelbox states.
Downsides
- Jira alone is often priced only as context for code.
- Backups leave out automation flows, app data and Assets.
- Exclusivity windows block other buyers: Turing's 90 days, and Labelbox's exclusive default.
Alternatives
- License a subset: one project, closed issues or older years.
- License Jira history together with the codebase, see selling or licensing source code and codebases.
- Keep the data and change contribution settings where the plan allows.
- Do not license at all.
Jira Among the Other Source Systems and the Seller Steps
Jira is one of the workplace systems whose export rights decide what a company can license, next to chat, email, wikis and code hosts. The parent page gives export rights and mechanics by source system, including how Jira compares with the others.
Sibling pages cover chat history in Slack data and AI training and Microsoft Teams data export, and the same Atlassian family appears in Confluence data, Bitbucket export and Linear issue export.
The seller steps that touch Jira run in this order:
- Run a rights audit of the Jira projects.
- Build the data inventory of projects, years and size.
- Check customer consent and confidentiality clauses.
- Apply employee data and AI training rules by jurisdiction.
- Write the data manifest, including the backup exclusions.
- Plan delivering licensed data.
- Decide should you sell or license your data at all.
EU Customers: Data Act Switching Is Not a Sale Right
The EU Data Act's switching chapter (Regulation (EU) 2023/2854, Chapter VI) gives cloud customers a route to export data when they change provider, but it is not a general right to export data for sale. The Regulation applies since September 12, 2025, and switching charges are removed from January 12, 2027; see EU Data Act and AI training data. This is information, not legal advice.
Jira Data From a Closed Company
A company that is closing should export Jira before its subscription ends, because Atlassian restores only from backups no older than 30 days. The steps are in selling data from a defunct company, and deadlines at vendors are covered in exporting company data before shutting down.
Buyer Programs Cited on This Page
The program terms above come from each program's own pages, checked November 9, 2026. Programs cited: Avelence, Centific, Collinear AI, Corpus Bridge, DataVendor (HUD), Labelbox Prism, Mercor, Ooak Data, Scale AI, Troveo and Turing Project Lazarus. See company data partnership programs, the Mercor data buyer profile and who buys AI training data.
Frequently Asked Questions
How do I export all data from Jira Cloud?
A Jira Cloud backup is the fullest export, but it leaves out items such as automation flows, third-party app data and Assets. It can be created every 48 hours with attachments, and the restore limit is 30 days (checked November 9, 2026).
Does Atlassian use Jira data to train AI?
Since August 17, 2026, Atlassian draws on customers' metadata and in-app data for its products and AI, with opt-outs that depend on the plan. The settings are listed in Table 2. The platform-wide how-to is in opting out of AI training.
Can a company sell Jira data to AI companies?
A company can license Jira data it controls once personal data, customer terms and secrets are dealt with, and several programs ask for it. Scale AI states it anonymizes inside the company's environment (checked November 9, 2026). The limits are in what company data cannot be licensed.
Do Jira backups include Marketplace app data?
No; third-party app data is excluded from Jira Cloud backups, and Marketplace developers may not sell End User Data. The exclusion is in the backup documentation, and the ban is s. 9.2 of the Developer Terms.