Atlassian has switched its AI training data collection to opt-out rather than opt-in across its suite of products, which includes Jira, Confluence, Trello, and Bitbucket. Unless an admin goes in and changes the settings, customer data is now being used to train Atlassian's AI models.
That's a meaningful shift for enterprise teams. Project management software holds some of the most sensitive internal data a company has - sprint plans, bug reports, internal documentation, customer-facing roadmaps. Most employees writing Jira tickets or Confluence pages aren't thinking about whether that content ends up in a training dataset.
What Admins Need to Do
Atlassian does provide a way to opt out, but the burden is on workspace administrators to find it and flip the switch. If you're on a company Jira or Confluence instance and nobody has checked the privacy settings recently, your organization is likely opted in already.
The opt-out path sits inside the admin console under privacy or data settings - the exact location varies by product. Admins managing multiple Atlassian tools will need to check each one separately.
A Pattern Worth Watching
Atlassian isn't doing anything unusual here - this is now a standard move across SaaS platforms rolling out AI features. Companies flip the default to "on" because adoption numbers look better, and most users never change defaults. Microsoft, Google, and Zoom have all navigated similar controversies with their own AI training policies in the past two years.
The difference with Atlassian is that its tools sit deep inside enterprise workflows, often at companies that have strict data governance requirements. Regulated industries - finance, healthcare, legal - should treat this as a compliance issue, not just a privacy preference. An opt-in model would have been the more respectful approach for a product used primarily by businesses.