Atlassian enables default data collection for AI training

Starting August 17, 2026, Atlassian will collect metadata and in-app content from Atlassian Cloud products such as Jira and Confluence by default, using it to train AI features like Rovo and Rovo Dev.

This applies to approximately 300,000 customers, with defaults varying by plan.

• Free / Standard: Metadata collected by default, in-app data also on by default
• Premium: Metadata collected by default, in-app data off by default
• Enterprise: Both off by default, metadata can be opted out

The metadata collected includes readability scores, complexity scores, task classification, semantic similarity metrics, story points, sprint end dates, and Jira Service Management SLA values.

In-app data includes Confluence page titles and body content, Jira issue titles, descriptions, and comments, custom emoji names, custom status names, and workflow names.

Atlassian stated that direct identifiers are removed, and the data is aggregated and protected before being used for training. The retention period is up to 7 years, and after deletion or opt-out, in-app data is removed within 30 days, with models trained on that data retrained within 90 days to remove its contribution.

However, the following are excluded from collection:

• Customers using customer-managed keys
• Atlassian Government Cloud
• Atlassian Isolated Cloud
• Customers with HIPAA requirements

This change marks a shift from Atlassian's earlier policy of "not using customer data for AI training," and is likely to reignite debate over data governance and model provenance in enterprise SaaS.