Atlassian’s new data collection policy protects rich customers while AI eats the rest
From August 17, the outfit will collect customer metadata by default unless you pay for the top tier
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.
Beginning in August, the company will seek to collect two types of data from its 300,000 global customers: metadata and in-app data from Jira, Confluence, and its other cloud products, which will then be fed into the company's models.
Metadata includes readability scores and complexity ratings for Confluence page content, task classifications assigned to content (such as "sales work item"), semantic similarity scores measuring how similar two Confluence pages are, and numbers entered into Atlassian-created fields – specifically story points assigned to a Jira work item, the end date of a sprint in Jira, and the Service Level Agreement of a Jira Service Management request.
For the metadata collection, lower-paying customers are on the hook no matter what. “If an Atlassian customer's highest active plan is Free, Standard, or Premium, metadata contribution is always on, and they are not able to opt out,” Arseny Tseytlin, head of product communications at Atlassian, told The Register via email. “All metadata is de-identified and aggregated before it is used to improve apps and experiences for all customers. We remove information that directly identifies individuals, such as name and email addresses.”
Once it collects the information, Atlassian says it will store it for up to seven years.
“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,” Tseytlin said. “By unlocking deeper insights into customer behavior, we’re able to drive continual improvements to (the) overall experience.”
Atlassian said that the training will give its apps the ability to surface the most relevant results in response to prompts and queries, summarize relevant content more accurately and concisely, identify the best templates for creating new documents, and learn which agentic workflows and follow-up questions lead to successful completions, making multi-step tasks faster and less confusing.
In-app data would be the content created by users within the Atlassian platform such as titles and content in Confluence pages, titles, descriptions, and comments in Jira work items, custom emoji names, custom Jira or Confluence status names, and custom workflow names. The Free and Standard tiers have this data collection turned on by default but can opt out; Premium and Enterprise customers have it turned off by default.
Tseytlin said that some Atlassian customers are completely excluded from metadata or in‑app "data contribution" entirely. This includes those who use customer-managed keys, or bring your own key, Atlassian Government Cloud, or Atlassian Isolated Cloud users. He said Atlassian will also not collect metadata or in-app data from customers with HIPAA compliance requirements or from some government and financial services customers.
The plan goes into effect on August 17.
“If customers were to right now terminate their contract, the new data contribution settings will not apply to them as these will not be enforced until August 17, 2026,” Tseytlin said.
After that date, once a customer opts out of having their data collected, or deletes their Atlassian apps, Atlassian will remove the corresponding in-app data from its datasets within 30 days. He said that, within 90 days, Atlassian will re-train any models previously trained on that data. ®