Atlassian has updated its data practices to allow training AI models on customer-generated content from Jira and Confluence, effective August 17, 2026. For the millions of businesses that run their operations on Atlassian’s tools, the change raises real questions about who owns the benefit when your work becomes training data.

What Is Actually Changing

The policy update means Atlassian can use in-app data — the actual content inside your tools — to improve its AI products. That includes Confluence page titles and body content, Jira issue titles, descriptions, and comments, custom workflow names, and custom field names.

This goes further than typical product analytics. Atlassian has previously used telemetry and usage patterns to improve its products. Now it is moving into the content layer: the projects you manage, the documentation your teams write, the processes your business runs.

The data is de-identified and aggregated before use, and Atlassian says it will not include content from customers with HIPAA compliance requirements, or from some government and financial services customers.

The Default Settings Matter

How the policy applies depends on which tier you are on:

• Free and Standard plans: In-app data collection is enabled by default. Customers can opt out.
• Premium and Enterprise plans: In-app data collection is disabled by default. Customers must actively opt in.
• Metadata collection (usage data, behaviour patterns): Enabled by default for all plans. Only Enterprise customers can opt out.

This means most Atlassian customers — the majority of whom are on Free or Standard plans — will be opted in automatically unless they take action before the August deadline. Free and Standard plan users who want to opt out need to find the setting and turn it off before August 17.

Atlassian has also committed to a meaningful data removal process: if a customer opts out or deletes their apps, Atlassian will remove the corresponding data from its datasets within 30 days, and retrain any affected models within 90 days.

Why Atlassian Is Doing This

The calculus for Atlassian is straightforward. The company’s AI product, Rovo, competes with Microsoft Copilot for business, GitHub Copilot for developers, and an expanding field of purpose-built AI agents for project management. Building better models requires better training data, and customer content is far richer than anything a lab can generate synthetically.

Atlassian is not alone in making this move. Microsoft has trained on enterprise usage data for years. Salesforce feeds customer interaction data into Einstein. The difference is that most of these companies have quietly done this at the metadata level. Moving into content-level training is a step change that more vendors will likely follow.

What This Means for Business

Review your plan tier now. If your business is on Free or Standard, you are opted in by default. Log into your Atlassian admin settings and review the AI data contribution toggles before August 17. If your work contains commercially sensitive information — and most business Jira and Confluence instances do — you should make an active decision rather than letting the default decide for you.

Understand what “de-identified” actually means. Atlassian says it aggregates and de-identifies data before training. But de-identification is not the same as anonymisation, and it is worth asking your IT or legal team whether the data in your Confluence instance would be protected adequately under Atlassian’s definition.

This is a policy, not just a checkbox. The real issue is not whether one vendor trains on your data. It is whether your business has a clear position on what AI vendors can and cannot do with the content your teams produce. Most businesses do not have that position written down. They should.

Know the difference between shared and dedicated AI. Many AI tools — including agentic platforms and enterprise AI advisors — learn from your data to improve their own general models, which benefits other customers too. Others, including custom-built AI solutions, can be structured so that your data trains models that work exclusively for you. If your business is deploying AI across core operations, understanding this distinction is worth the hour it takes to ask vendors directly.

The Atlassian move is a signal that AI data collection is moving from the periphery to the mainstream of enterprise software. The businesses that handle it well are the ones that have policies in place before the defaults kick in, not after.

---

Enterprise DNA put together a free field guide on exactly this: the full Claude ecosystem, Claude Code, and how to roll agents out without breaking things. Get the guide.