As AI becomes embedded in analytics workflows, K–12 data teams are asking:
“If we use Fabric Copilot, where is our student data actually processed?”
This matters because under NSW PPIPA and HRIPA, even temporary processing of student data outside Australia can count as a disclosure.
Here’s the current state of Fabric Copilot for Australian tenants.
🟦 What’s Safe Today (100% Onshore)
These Copilot features are fully deployed in Australian datacenters and DO NOT require cross-geo processing:
✔ Copilot in Dataflows (Data Factory)
- Transformation suggestions
- Expression help
- Data cleaning prompts
✔ Copilot in Notebooks (Synapse)
- Code generation (SQL/PySpark)
- Data exploration
- Transformation recommendations
👉 All AI inference stays in Australia.
🚫 What’s NOT In-Region Yet (Requires Cross-Geo)
These features require AI inference outside Australia and therefore cannot run with cross-geo disabled:
❌ Power BI Copilot (built-in)
- Natural-language to report
- Q&A exploration
- Narrative summaries
❌ Standalone Copilot Experiences for Power BI
❌ Semantic Model Copilot
👉 With cross-geo OFF (recommended for schools), these features simply stay disabled — no data leaves Australia.
🔧 The Critical Setting
To stay compliant:
Leave cross-geo AI processing OFF.
If you do, all supported Copilot features will only call Australian Azure OpenAI endpoints.
🛡 Data Schola Guidance for Schools
- Use Copilot only in Dataflows and Notebooks
- Keep cross-geo processing disabled
- Avoid Copilot for sensitive datasets (behavioural, wellbeing, medical, disability)
- Start preparing data now for future Power BI Copilot availability
Full support for Power BI Copilot in Australia is expected as Fabric approaches General Availability.
👉 Start preparing your data today to be AI-ready when that moment arrives.
