DataPacts

Confluence Data: Export and Licensing Wikis for AI

Checked

At a glance

A company can license its Confluence wiki for AI training because Atlassian's Customer Agreement (effective October 1, 2026) leaves ownership of Customer Data with the customer, but the content must first leave Atlassian Cloud, and each export route leaves something out.

Independent reference. Every buyer is covered on the same terms, every figure carries its source and check date, and no buyer pays to be listed or placed higher.

Request an independent introduction

Tell us what data your company holds. We reply with the buyers that fit and what their terms leave out. Your details reach a buyer only after you agree to a named introduction.

IndependentYour details stay with us until you agreeNot selling is a valid outcome

A site backup link stays available for 14 days, and a site admin's CSV or XML export also carries restricted pages. Four programs name Confluence among the systems they want (checked November 9, 2026).

Confluence is one source in AI training data licensing, and the choice is the company's: licensing selected spaces, licensing nothing and keeping the wiki internal are all valid results.

Fact-check note: the Atlassian terms and documentation below were read on November 16, 2026, and program terms are the programs' own statements; how we source, date and check every figure is set out in the methodology.

What Licensing Confluence Data for AI Training Involves

Licensing Confluence data means granting an AI developer, or a vendor or intermediary that supplies one, the right to train on a defined copy of the company's spaces, pages and attachments, while the company keeps ownership.

Atlassian is the host of the wiki, not a party to that license, and the license covers a copy that the company has exported or made readable, not Atlassian's platform.

Programs describe the grant in different ways. SuperAnnotate states that "You're licensing rights for a set period, not selling the data" (checked November 9, 2026), while micro1's Data License Master Terms, updated November 8, 2026, describe a nonexclusive, irrevocable, perpetual, sublicensable license. Centific, Mercor, Scale AI and Turing each state their own route (table below).

The data type itself, including per-page figures, trade secrets and third-party content in wikis, is covered on the page about selling documents, wikis and drive files for AI training. This page covers what is specific to Confluence: Atlassian's terms, the export routes and the admin exception.

What Atlassian's Terms Say About Customer Data

Two Atlassian documents govern customer content in Confluence Cloud: the Atlassian Customer Agreement, effective October 1, 2026, and the Atlassian AI Terms, effective August 17, 2026. The table compares the clauses that matter for licensing, as of November 16, 2026, with Atlassian's own documents as the source.

Term What it says Date Source
Customer Agreement, definition Customer Data includes content provided by the customer or its users, including from Third-Party Products effective 2026-10-01 atlassian.com/legal/atlassian-customer-agreement
Customer Agreement s.13 Customer owns Customer Data as between the parties effective 2026-10-01 atlassian.com/legal/atlassian-customer-agreement
Customer Agreement s.4.1, s.4.4 Processing per the data processing addendum; retrieval per the Documentation effective 2026-10-01 atlassian.com/legal/atlassian-customer-agreement
Customer Agreement s.12.4 Deletion after expiration or termination per the Documentation effective 2026-10-01 atlassian.com/legal/atlassian-customer-agreement
AI Terms s.2.1 Input and Output are Customer Data; no subcontractor training effective 2026-08-17 atlassian.com/legal/ai-terms
Data contribution Atlassian's own use of metadata and in-app data; plan-dependent defaults in effect 2026-08-17 atlassian.com/trust/ai/data-contribution
Developer Terms s.9.2 Marketplace developers "may not sell any End User Data" effective 2025-12-01 developer.atlassian.com

Ownership Under Section 13 of the Customer Agreement

Section 13 of the Customer Agreement states that "As between the parties, Customer owns all intellectual property and other rights in Customer Data and Customer Materials provided to Atlassian or used with the Products", and Customer Data covers content provided by the customer or its users, including from third-party products. Section 2.2 restricts renting, leasing, selling or sublicensing the Products themselves, not Customer Data.

The agreement contains no clause on record that permits or forbids licensing Customer Data to a third party for AI training, so this page states only what section 13 says. Ownership against the host is one layer; authors, clients and contractors are others, and a rights audit for data you want to license covers them.

Processing, Retrieval and Deletion Under the Customer Agreement

The Customer Agreement lets Atlassian process Customer Data only as set out in the data processing addendum (s.4.1), points to the Documentation for how a customer retrieves its data (s.4.4), and says Atlassian deletes Customer Data after expiry or termination in accordance with the Documentation (s.12.4).

The Documentation sets the deletion period after cancellation, so a company should check it before cancelling a site.

Atlassian's AI Terms and Its Own Data Contribution

Atlassian's AI Terms treat what users type into Atlassian's AI features, and what those features return, as Customer Data, and section 2.1 bars Atlassian's subcontractors from using that Input or Output to train or improve their models. The AI Terms supplement the Customer Agreement and control on conflict.

Atlassian's separate data contribution for its own products and AI has been in effect since August 17, 2026, with defaults and opt-outs that depend on the plan; it is the vendor's own use, not a license route and not a payment.

The plan-by-plan settings are set out on the Jira page linked in the contextual border below.

Marketplace Apps Reading Confluence

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 Confluence is not a route for its developer to resell wiki content.

A company licensing its own wiki is not the app developer, and access granted to a program's app is a separate question covered under buyer connectors vs native exports.

How Confluence Content Leaves Atlassian Cloud

Confluence content reaches an AI buyer as a page or space export, as a site backup, or through a connection or in-place process that a program runs, and each route leaves different content behind. The table lists the routes as of November 16, 2026; the source is Atlassian documentation unless a program is named.

Route Who runs it What it carries or leaves out Source
Page or blog post export (Word, PDF) any user who can see it published version only; no comments in PDF support.atlassian.com
Space export (PDF, CSV, HTML, XML) space admin only content visible to the exporter; CSV includes comments and attachments by default; blog posts not in PDF or HTML support.atlassian.com
Admin CSV or XML export site admin all content, restricted pages included support.atlassian.com
Site backup (Backup manager) admin one file kept; link 14 days; up to 30 GB plus 800 GB attachments; no soft-deleted spaces; databases not exported support.atlassian.com
Read-only connection program (Centific) in-scope systems connected read-only centific.com, checked 2026-11-09
In-place anonymization program (Scale AI) anonymized inside the company's environment; per-package approval scale.com, checked 2026-11-09

Page and Space Exports: Formats and What They Leave Out

A single page or blog post exports to Word or PDF, and a space admin can export a whole space, or selected pages, to PDF, CSV, HTML or XML (Atlassian documentation, checked November 16, 2026). Atlassian's page lists what each format omits:

  • Comments are never included when exporting to PDF, and HTML exports omit page comments; CSV includes comments and attachments by default.
  • Only the published version of a page is exported.
  • Blog posts are not included in space PDF or HTML exports.
  • Team Calendars are not exported this way.

An HTML export places attachments in a folder per page ID, so a buyer receiving that format gets files that must be matched back to pages. The documentation read does not describe page history, and this page states nothing about it.

A Confluence Cloud site backup is created in Backup manager, only one backup file is stored at a time, and the link is available for 14 days.

A backup can hold up to 30 GB plus 800 GB of attachments, and Atlassian states that soft-deleted spaces are not backed up and that databases cannot currently be exported. Atlassian's own backups, taken every 24 hours, serve application recovery only.

The backup is a restore file, so a buyer receives a derived set, and the exclusions belong in the data manifest as stated omissions.

Read-Only Connections and In-Place Anonymization by Programs

Some programs read Confluence in place instead of receiving an export: Centific states that in-scope systems are connected read-only, and Scale AI states that it anonymizes inside the company's environment before extraction, with per-package approval (both checked November 9, 2026).

Mercor states that no AI lab buyer or third-party vendor sees the raw data, and Turing takes samples through a private folder.

No scope facts are on record for what an app reads, so a company should ask in writing which spaces an app reads and when access ends.

Restricted Pages, Admin Exports and Personal Data in a Confluence Wiki

A Confluence export contains only content visible to the person who runs it, except CSV and XML exports by a site admin, which include all content, so pages restricted to HR, legal or management leave with an admin export unless they are removed first.

The same visibility rule means that an ordinary space admin's export silently omits pages the admin cannot see, which is why exports differ between runs by different people.

Wikis also hold personal data in places a text search does not find, and PII removal and de-identification describes the methods. The items to check before any export leaves the company:

  • restricted spaces and pages
  • user names, mentions and profile links
  • comments and inline comments
  • attachments such as screenshots, logs and exports
  • pasted credentials and customer details

Space and site admins can export content owned by deactivated accounts, and organization admins can block exports for certain products or classification levels, which helps keep restricted spaces out of any route.

Employees named in pages are a notice question, covered under employee notice and consent.

What Buyers Want From Wikis and Which Programs Name Confluence

Four programs on record name Confluence among the systems they want, each listed here with its role and the date its page was checked. The table is as of November 9, 2026 (Turing's data list re-read November 16, 2026), alphabetical, with the program's own page as the source and no payout column by design.

Program Role Where Confluence appears (program's wording) Route or term relevant to wiki data Checked Source
Centific Enterprise Data data vendor "Office files, Drive, SharePoint, Confluence, Notion, Jira, Asana" among 188 supported apps read-only connection; "You license your data. You never hand it over."; mutual NDA 2026-11-09 centific.com/enterprise-data
Mercor Data Monetization human-data platform "Confluence" among 100+ supported tools; Software Company listing names Confluence covers extraction and anonymization costs; raw data not seen by lab buyers 2026-11-09 mercor.com/data; api.data.mercor.com
Scale Data Partnerships human-data platform "Jira, Confluence, Asana, Notion, Linear" anonymizes inside the company's environment; exclusivity, term and deletion not published 2026-11-09 scale.com/data-partnership
Turing Project Lazarus human-data platform; resells the record on the company's behalf "Wiki and docs": "Notion, Confluence, Drive" license to Turing exclusive for the first 6 months; Turing may license the record to more than one lab in that term (re-read 2026-11-17); covers de-identification costs 2026-11-09 (data list re-read 2026-11-16) lazarus.turing.com

Programs are listed alphabetically; the list implies no ranking or endorsement.

Other programs name wikis or knowledge bases without naming Confluence. SuperAnnotate names "process docs, runbooks, wikis" and states that the seller approves each buyer. micro1 names customer support tickets and knowledge bases. OpenTrain AI, a broker that states it never receives or holds data, names documents and wikis (all checked November 9, 2026).

Exclusivity terms differ: Turing's 6-month exclusive term (table above) still lets it license the record to more than one lab, and Scale AI has not published its exclusivity terms. A fuller comparison sits in AI data partnership payouts and program terms compared.

Licensing Confluence Content: Benefits, Risks, Downsides and Alternatives

Licensing a Confluence wiki can earn income from records the company already keeps without giving up ownership, and it also releases text that cannot be recalled once a model has trained on it.

Benefits

  • The company keeps ownership under Customer Agreement s.13, and Turing states that ownership does not change hands.
  • Some programs carry preparation costs: Mercor covers extraction and anonymization, and Turing covers de-identification (both checked November 9, 2026).

Risks

  • Restricted pages travel in an admin CSV or XML export.
  • Wikis hold personal data that must be found and handled before delivery.
  • Labelbox states that "Models already trained on the data cannot be unlearned" (checked November 9, 2026).

Downsides

  • Every route leaves content out: comments, blog posts, unpublished versions or soft-deleted spaces.
  • Turing's 6-month exclusive license limits the company's own selling during that period.

Alternatives

  • Not licensing: keep the wiki internal.
  • License selected spaces only.
  • License pages without comments or attachments.
  • Use the content for internal AI tools only.

The full weighing is on the page about should you sell or license your data, and the right outcome for a given wiki can be to keep it.

Where Confluence Fits Among Source Systems and the Seller Steps

Confluence is one of several workplace systems a company may license, and its export comes after the inventory and before delivery.

The comparison across systems is export rights and mechanics by source system, Atlassian's contribution plan table sits on the page about Jira data export rights, SharePoint wikis are covered in Microsoft 365 exports for data licensing, and Drive documents in Google Workspace exports.

Atlassian's other products have their own pages: Notion workspace export is a related wiki tool, and Bitbucket export covers repositories.

  1. Build a data inventory.
  2. Check customer consent and confidentiality clauses.
  3. Plan security during an AI data sale.
  4. Start choosing a buyer for your data.
  5. Negotiate the AI data licensing agreement.
  6. Plan delivering licensed data to an AI buyer.

Confluence Content of a Closed or Insolvent Company

A closed company can license its wiki only if someone with authority still controls the Atlassian site, and the Customer Agreement leaves the deletion period after termination to Atlassian's Documentation.

The steps for such a company are set out in selling data from a shut-down or defunct company. Vendor deletion clocks are covered in exporting company data before shutting down, and Turing states that it accepts companies shut down within or beyond 90 days.

Program Pages Named on This Page

The program terms above come from each program's own page, checked on the dates in the table.

Frequently Asked Questions

How do I export a complete Confluence space?

A space admin exports a whole space to PDF, CSV, HTML or XML from the space settings, and only CSV or XML exports run by a site admin include content the exporter cannot see. Comments and blog posts are left out of some formats (see the export table above).

Can a company sell its Confluence wiki to AI companies?

A company can license wiki content it owns, usually non-exclusively, through data programs and intermediaries, once restricted pages, personal data and client material are dealt with.

Which company data cannot be licensed at all is set out under what company data cannot be licensed.

Does Atlassian train AI on Confluence content?

Atlassian's data contribution for its own products and AI has been in effect since August 17, 2026 with plan-dependent defaults, and its AI Terms bar subcontractors from training on AI feature Input and Output.

Platform settings and their limits are covered under opting out of AI training.

What is a Confluence wiki worth to an AI buyer?

No public price exists for Confluence data as such; Mercor's software-company listing, which names Confluence among its tools, states USD 50,000 to 500,000 to the company for code-led datasets (published October 17, 2026, checked November 9, 2026).

That range is Mercor's claim, not a payment, and the valuation methods are covered under how much company data is worth to AI companies.

Is Confluence being phased out?

Atlassian announced on September 8, 2025 that its Data Center products stop selling to new customers on March 30, 2026 and reach end of life on March 28, 2029; nothing on record ends Confluence Cloud. Export mechanics for self-managed sites are not covered on this page.