Atlassian Enables Default Data Collection to Train AI

Atlassian just flipped the switch: starting August 17, 2026, your Jira and Confluence activity will help train its AI by default. The company says it’ll use both usage signals (“metadata”) and some in‑app text like titles and comments to make search and summaries smarter, and it may keep contributions for up to seven years. The fine print: smaller plans can’t turn off usage signals, in‑app text is on by default for Free/Standard but can be toggled, Premium starts in‑app off, and only Enterprise can fully say no. Some groups (government, HIPAA, special encryption) are excluded. Atlassian promises it strips names and IDs, removes in‑app data within 30 days, and “untrains” models within 90 days after opt‑out. Docs are here: Atlassian data contribution settings.

Cue the meltdown. One camp warns “it’s all your pages and tickets,” linking docs and shouting opt‑out confusion, while long‑time users vent that Atlassian “goes from misstep to misstep.” Another camp shrugs: data helps everyone, calling it a “genius move” and comparing it to sharing knowledge on Stack Overflow. Conspiracy corner? Rumors swirl that an Anthropic deal needs a “clean” business dataset. The memes write themselves: “My backlog finally useful,” “AI trained on Jira chaos,” and “the seven‑year itch… for your tasks.” The spiciest drama: small teams feel forced to trade privacy for price, while big enterprises get the off switch. Classic internet split—privacy panic vs progress hype—with extra heat over who gets control.