Starting today, Atlassian begins collecting metadata and in-app content from Jira, Confluence and its other cloud products by default to train and improve its AI features — an opt-out model that has drawn sharp criticism from customers and a rare counter-campaign from a competitor.
Under the policy, which Atlassian detailed on its "Data practices built for responsible AI" page, the company is updating how it uses customer data to "deliver improved AI experiences for all". Administrators can disable the contribution in Administration → Security → Data contribution, but the default is on: customers who do nothing will have their metadata — and, depending on plan, in-app content such as pages and issues — fed into Atlassian's AI efforts, including its Rovo assistant.
The timing is notable: Atlassian frames the change as part of a broader push into AI-enabled knowledge work, with Rovo positioned as the connective layer across Jira, Confluence and other tools. For many enterprise customers, however, the shift raises governance questions: internal documentation often contains sensitive product plans, customer data and employee information, and default-on training has become a flashpoint in enterprise software.
Competitors have moved to capitalize on the unease. GitLab published a guide — "Atlassian will train on your data. Opt out with GitLab" — urging teams to evaluate alternatives, while third-party vendors have rushed out opt-out how-tos for administrators before the change takes effect.
Atlassian says the data is used to improve AI features for all customers, and that customers on affected plans can opt out at any time. The company has also published a FAQ covering what is collected, how it is protected and how to manage contribution settings.
The controversy mirrors a broader industry moment: as AI features become default in productivity suites, the boundary between a customer's data and a vendor's training corpus is being redrawn — often with the customer responsible for opting out, not the vendor for opting in.
