If you run Jira or Confluence Cloud, Atlassian is about to start using your data to train AI models — and depending on which plan you’re on, you may not get a say in it.
What Atlassian announced
Under Atlassian’s updated data policy, two categories of data get collected: metadata (readability scores, task classifications, story points) and in-app data (actual page content, issue descriptions, comments). How much control you have over that depends entirely on your subscription tier:
| Plan | Metadata collection | In-app data (content, comments) |
|---|---|---|
| Free / Standard | Mandatory, no opt-out | On by default — opt-out available |
| Premium | Mandatory, no opt-out | Off by default |
| Enterprise | Opt-out available | Off by default |
Effective August 17, 2026. Data is retained for up to seven years; opting out removes in-app data within 30 days and affected models are retrained within 90. Source: gHacks Tech News.
In plain terms: unless you’re paying for Enterprise, you cannot fully opt out of having your workspace data feed Atlassian’s AI training. Free and Standard customers — likely the largest slice of Atlassian’s roughly 300,000 customers — get the least control by default.
The reaction hasn’t been subtle. GitLab published a direct response offering an opt-out path by switching platforms — and customers have been just as blunt:
“These incremental bait & switches are despicable.”
— Jason M., Atlassian Community forum“This is such an obvious conflict of interest … violating their client’s trust.”
— josefritzishere, Hacker News“Absolutely insane that this is legal. The only reason to do this is to trick and abuse customers.”
— rurp, Hacker NewsWhere GForge stands
We’re not going to pretend we have a clever policy toggle that Atlassian doesn’t. We have something simpler: GForge doesn’t have a model to train. We don’t build or operate a proprietary AI model, so there’s no training pipeline for your data to feed — for any customer, on any plan.
No model to train
GForge has no proprietary LLM. There is no training pipeline that touches your tickets, wiki pages, or code — for anyone, at any tier.
AI is opt-in, every time
AI integrations (Flow AI, MCP-based tools) are off until you or your admin turns them on, per project. Nothing is enabled by default.
Your AI vendor, your terms
When you connect an AI tool, that tool’s own data and training policy governs what happens to what you send it — not GForge’s. We connect you to it; we don’t sit in the middle collecting a copy.
Questions we expect you’ll ask
Does GForge train AI models on my data?
No. GForge doesn’t build or operate its own AI model, so there’s nothing for your data to train. This isn’t a policy we could quietly change in a future terms update — it’s a fact about how the product is built today.
I’m an existing customer — do I need to change any settings?
No action needed. This isn’t a new policy we’re rolling out in response to Atlassian’s news — it reflects how GForge already works.
What happens if I enable Flow AI or connect an MCP-based AI tool?
You’re choosing to send data to that specific AI product, and that product’s own data-handling and training policy applies to it — not GForge’s. We’d recommend checking your chosen provider’s terms before enabling any integration, the same way you would for any third-party tool.
Is this the same as Atlassian’s “opt-out”?
No — there’s nothing to opt out of. Atlassian’s opt-out (where available) stops your data from being used by a training pipeline that exists and runs by default. With GForge, that pipeline doesn’t exist in the first place.
Not ready to have your backlog train someone else’s model?
See how GForge compares to Jira and Confluence, or talk to us about migrating your projects over.