Atlassian AI policy now active; Jira, Confluence admins get new controls

Atlassian AI policy now active; Jira, Confluence admins get new controls

August 17, 2026

SAN FRANCISCO, August 17, 2026, 10:05 PDT

  • Atlassian started new customer-data contribution terms on Monday.
  • All cloud plans have the option to turn off in-app content contribution.
  • Turning off metadata contribution is only available to Enterprise organizations.

Atlassian Corporation (NASDAQ:TEAM) rolled out new customer-data contribution settings starting Monday. The update lets eligible metadata and content from user activity help boost products and AI features for other customers.

Control is the main difference. Organization admins on all cloud plans have the option to reject in-app data sharing. But just Enterprise accounts can also switch off metadata contribution.

So now, three of Atlassian’s four plan tiers don’t offer a way to opt out of metadata. Free and Standard start with in-app contributions turned on. Premium and Enterprise launch with it off.

Top cloud planDefault metadataMetadata opt-outDefault in-app dataIn-app data opt-out
FreeEnabledNoEnabledAllowed
StandardEnabledNoEnabledAllowed
PremiumEnabledNoDisabledAllowed
EnterpriseEnabledAllowedDisabledAllowed

The top active plan in each cloud org sets the defaults, even for trials. Admins can check these in Atlassian Administration under Security, then Data contribution. Each organization needs to be reviewed on its own.

Metadata covers things like stats and regular patterns. These can be Jira story points, how complex a Confluence page is, or what topics people search for a lot in Rovo. Atlassian says it strips out names and emails and puts the data together.

In-app data is straightforward. This covers Confluence page content, Jira titles, descriptions and comments. It also includes custom workflows, status names and emoji names.

Data categoryVerified examplesCompany safeguardsAdmin control
Metadata: content attributesStory points, page complexity, sprint dates, service levelsDe-identified, only shared in aggregateEnterprise only
Metadata: common patternsTop search terms, Rovo Chat subjects, config patternsLeaves out unusual org-level patternsEnterprise only
In-app dataPage content, work-item titles, descriptions, comments, custom labelsDe-identified, aggregated with others before cross-customer useAll plans

Arseny Tseytlin, who runs product communications at Atlassian, said lower-tier metadata contribution is “always on” and customers in those tiers don’t have an opt out. The Register

The first group includes Jira, Confluence and Jira Service Management, as well as platform apps like Rovo, Assets and Analytics. Products like Loom, Trello, Bitbucket, Data Center offerings and Marketplace apps aren’t part of it yet.

Product or organizationCurrent statusImportant detail
Jira, Confluence, Jira Service ManagementIncludedSettings are set at the organization level
Rovo and other platform appsIncludedSame organization settings manage these
Loom, Trello, BitbucketNot included yetAtlassian says it will give a heads up before any changes
Data Center and Marketplace appsNot includedNo data is shared as part of this change
Government, Isolated Cloud, HIPAA, CMK or BYOKExcludedContribution can’t be switched on for these

Atlassian said customer data that users provide isn’t used to train language models hosted by third parties. Those vendors work under zero-data-retention deals. Atlassian may fine-tune open-source models on its own systems using certain metadata.

Switching off contribution doesn’t remove data right away. Atlassian says in-app data will be taken out of improvement datasets within 30 days. Metadata goes within 90 days, and models using that data get retrained.

Milestone or actionTimingStatus
Started rolling out settingsApril 16, 2026Done
Settings rollout wrapped upMay 19, 2026Done
New terms and data policy in forceAugust 17, 2026Active
Remove in-app data if user opts out or deletesWithin 30 daysPledged
Remove metadata after user opts out or deletesWithin 90 daysPledged

Risks: The company says it controls safeguards and data removal, but that’s their own claim. Customers have to judge for themselves if de-identification works for their security. If an admin leaves a setting unchecked, eligible content could be used in product improvement across customers until someone updates it.

The immediate work is on admin steps. Free and Standard users need to check the in-app toggle. Each organization should keep a record of which plan manages its defaults.

BEZ KABLI • EXTENDED COVERAGE

Further analysis

What changed in Atlassian's data policy on August 17, 2026?
Atlassian began using eligible customer metadata and in-app data to improve apps and AI experiences across customers. The first phase covers Jira, Confluence, Jira Service Management, platform apps such as Rovo and certain Teamwork Graph connectors.
Can an organization opt out?
Every cloud plan can disable in-app data contribution. Only Enterprise organizations can disable metadata contribution. Metadata remains mandatory on Free, Standard and Premium plans, except for organizations excluded by compliance requirements.
Which plans contribute in-app data by default?
Free and Standard plans default to On. Premium and Enterprise plans default to Off. An organization admin can change the in-app setting on every plan.
What data can be contributed?
Metadata can include story points, page complexity and common Rovo search topics. In-app data can include Confluence page bodies, Jira titles, descriptions, comments and custom workflow names. Atlassian says contributed data is de-identified and aggregated before cross-customer use.
Does Atlassian send this data to outside AI providers for training?
Atlassian says no. Its third-party-hosted language-model providers operate under zero-data-retention agreements. Atlassian may fine-tune open-source models inside its own infrastructure using eligible metadata.
How quickly is contributed data removed after an opt-out?
Atlassian promises to remove corresponding in-app data within 30 days and metadata within 90 days. It also says affected models will be retrained. These are company commitments rather than independently verified results.

Artur Ślesik

Artur Ślesik is a technology and financial markets journalist at Bez-kabli.pl, covering artificial intelligence, semiconductors, technology stocks and emerging innovations. A graduate of Warsaw University of Technology, he combines a technical background with market analysis to explain how new technologies are shaping industries, businesses and investment trends worldwide.