Atlassian Switches Customer Data AI Training Usage to Default On

Customer metadata and in-app content generated in Jira, Confluence, and other Atlassian Cloud products will be used by default for Rovo and Rovo Dev training starting August 17, 2026.

Defaults vary by plan as follows.

• Free / Standard / Premium: Metadata contribution is always enabled
• Free / Standard: In-app data contribution is enabled by default and can be changed
• Premium: In-app data contribution is disabled by default
• Enterprise: Both metadata and in-app data are disabled by default, retaining control

The data collected falls into two categories.

• Metadata: Readability/complexity scores, task classification, semantic similarity metrics, story points, sprint end dates, SLA values from Jira Service Management
• In-app data: Confluence page titles and body content, Jira issue titles/descriptions/comments, custom emoji names, custom status names, workflow names

Atlassian states that it applies removal of direct identifiers, data aggregation, and protective measures before training, and explains that contributed data is retained for up to 7 years. After deletion or opt-out, in-app data is removed within 30 days, and models trained on that data are retrained within 90 days to remove the contributed portion.

However, the following customer groups are excluded from data collection.

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

The key change is that business SaaS has moved more clearly toward using customer data for AI training by default, requiring enterprise customers to review this from a privacy, governance, and compliance standpoint.