Atlassian announced on April 18-19 that it will enable default AI training data collection across Jira, Confluence, and other cloud products starting August 17, 2026. The policy affects over 300,000 global customers and creates a controversial two-tier privacy system: Free and Standard tier customers cannot opt out of metadata collection, while Enterprise tier customers have both metadata and in-app data collection turned off by default. Data will be retained for up to 7 years to train Atlassian Intelligence features including Rovo AI, Chat, and automated agents.
This is privacy-by-paywall. Enterprise customers get full opt-out. Everyone else accepts data collection whether they like it or not.
Enterprise Gets Privacy, Others Don’t
Atlassian’s opt-out capabilities are tiered by subscription level. Free and Standard customers cannot opt out of metadata collection, which includes task classifications (“sales work item”), story points, sprint dates, readability scores, and semantic similarity measures. Premium users also cannot opt out of metadata but have in-app data collection OFF by default. Only Enterprise tier customers get both metadata and in-app data collection OFF by default with full opt-out controls.
The Register puts it bluntly: “Unless a customer pays for the most expensive enterprise license, or the law forbids it, Atlassian is going to collect their data to train its AI models, and you can’t fully opt out.”
For startups and small teams on Free or Standard plans, this means Atlassian will collect task classifications, story points, sprint velocity data, and semantic similarity scores without any ability to opt out. These metadata points can reveal competitive business strategies, team productivity trends, and project priorities. It’s exactly the kind of intelligence companies pay consultants to analyze. Now Atlassian gets it for free and uses it to train AI for all customers.
Metadata vs. In-App Data: What’s at Risk
Atlassian collects two types of data. Metadata includes readability scores and complexity ratings for Confluence page content, task classifications assigned to content, semantic similarity scores measuring how similar two Confluence pages are, story points assigned to Jira work items, sprint end dates in Jira, and Service Level Agreement values in Jira Service Management. This is MANDATORY for Free, Standard, and Premium tiers with no opt-out option.
In-app data is broader: Confluence page titles and body content, Jira work item titles, descriptions, comments, custom emoji names, custom status names, and custom workflow names. This data collection is ON by default for Free and Standard tiers (though admins can opt out), and OFF by default for Premium and Enterprise.
Atlassian claims metadata is “de-identified and aggregated” before use. However, project names like “Project Stealth Launch Q3” or workflow names like “Competitor Acquisition Review” reveal business strategy even without employee names attached. Moreover, story points plus sprint dates reveal team velocity. Furthermore, semantic similarity scores mean Atlassian’s AI compares your Confluence pages to other customers’ pages to find patterns—a cross-customer training dynamic that could inadvertently leak industry trends.
Related: GitHub Copilot Trains on Your Data: Opt Out Before April 24
Your 2026 Data Will Train AI Until 2033
Atlassian will retain collected data for up to 7 years. Even if customers opt out or delete their accounts, previously collected data may persist in training datasets for up to 7 years, with 30-90 day removal processing time for opt-outs. Atlassian justifies this with: “Retaining this data, which has been de-identified and aggregated at a customer level and is common across customers, enables us to make more meaningful observations over longer periods of time.”
Seven years is excessive for training data retention. A project archived in 2026 will remain in Atlassian’s AI training datasets until 2033. Consequently, for comparison, 7 years ago was 2019—imagine your 2019 project data (pre-pandemic workflows, outdated tech stacks, former team members’ comments) still being used to train AI in 2026.
Projects pivot, companies change strategy, employees leave, competitive landscapes shift. Nevertheless, Atlassian’s AI will continue learning from outdated project data for nearly a decade. Additionally, for regulated industries or companies with strict data retention policies, this creates a compliance nightmare. If you delete your Atlassian account in frustration, your data doesn’t disappear—it lives on in AI training datasets until 2033.
Opt-Out Options: A Tier-by-Tier Breakdown
Certain customers are automatically exempt from data collection: HIPAA-compliant accounts, customer-managed encryption key (BYOK) users, Atlassian Government Cloud users, Atlassian Isolated Cloud users, and some government and financial services customers. For everyone else, opt-out is available through “Data Contribution Settings” in the admin panel, rolled out on April 16, 2026.
Free and Standard tier users can only opt out of in-app data—metadata collection is mandatory. Meanwhile, Premium users get in-app data OFF by default but still cannot opt out of metadata. In contrast, Enterprise customers get both metadata and in-app data collection OFF by default with full opt-out controls.
Here’s the breakdown:
- Free: Metadata ON (cannot opt out) | In-app data ON by default (can opt out)
- Standard: Metadata ON (cannot opt out) | In-app data ON by default (can opt out)
- Premium: Metadata ON (cannot opt out) | In-app data OFF by default
- Enterprise: Metadata OFF by default | In-app data OFF by default
The 4-month notice period (April 16 to August 17) sounds generous, but many Atlassian admins won’t see this announcement or know where to find “Data Contribution Settings” in the admin panel. As a result, by August 17, countless teams will unknowingly opt into data collection simply by not opting out. For teams that want to protect their data but can’t afford Enterprise tier, the only option is to enable customer-managed encryption keys (BYOK)—a complex enterprise feature most small teams don’t use.
Key Takeaways
- Atlassian AI training data collection starts August 17, 2026 across Jira, Confluence, and other cloud products
- Free and Standard users cannot opt out of metadata collection (story points, task classifications, semantic similarity, sprint dates)
- 7-year retention means data collected in 2026 will train AI until 2033, even if you delete your account
- Enterprise tier is the only way to fully opt out of both metadata and in-app data collection
- Automatic exemptions apply to HIPAA accounts, Government Cloud, and customer-managed encryption key (BYOK) users
- Privacy-conscious teams should evaluate self-hosted alternatives like OpenProject and XWiki (70-80% cost savings vs. Atlassian Enterprise) or enable BYOK to auto-exempt from collection
