I ran into a worrying issue using the Atlassian Intelligence assistant to make very small, targeted edits to a Confluence Cloud page.
I asked the assistant to make surgical, content-only changes to a PRD (minor table edits, wording tweaks), explicitly not touching page structure or location.
The assistant described the planned changes and, after I clicked its confirmation prompt, responded with “Processed successfully.”
Immediately after refreshing the page:
At the same time:
From my perspective, this looks like:
An AI operation that:
Reports success,
Results in the page no longer being where it was in the space tree,
And leaves the assistant unable to surface a clear, auditable record of what structural changes (if any) it actually made.
Why I’m concerned:
I only requested content tweaks, but the page’s placement in the space changed around the same time as the AI action.
The assistant’s own “view” of the page state seems out of sync with what Confluence is actually showing me.
There’s no obvious, user-visible audit trail from the AI describing: “I changed these fields and also moved/archived this page.”