Where Does Your AI Data Actually Go?
Most organisations know where their files are stored. Far fewer know where the prompt goes when a staff member pastes a document into an AI tool, how long it is kept, who else can see it and under which country's law. Eight questions to find out which of three positions you are really in.
Answer for the AI tool your staff use most, not the one policy says they should use. If you do not know an answer, pick the option that says so. Not knowing is the finding.
AI Data Residency Quiz
Answer for the AI tool your staff actually use. Unknown scores zero.
A self-assessment to structure a review, not legal advice. Confirm every answer against the vendor's written terms. Nothing entered is recorded or transmitted.
The Three Positions
Data residency is not a yes or no. Storage, processing, retention and legal jurisdiction can each sit in different places, and the quiz separates them.
Sovereign
Data is stored and processed in Australia, on infrastructure you control or that is contractually bound to stay onshore, with no training on your content, known retention and no foreign subprocessor in the chain. This is what a private deployment in an Australian region or on-premises gives you.
Onshore stored, offshore processed
Files sit in an Australian region, but the model that reads them runs elsewhere, or the vendor reserves the right to route processing offshore. Storage location satisfies the letter of many policies while the prompt and the answer still leave the country.
Offshore
Data goes to a foreign region for processing, may be retained there, may be reviewed by the vendor or used to train models, and is subject to the law of that jurisdiction. Common with consumer and unmanaged tools, and often not what anyone intended.
How the Scoring Works
Eight questions, each scored zero to three, equal weight, maximum 24. The bands follow the three positions above.
Answer for the tool actually in use
The options run from fully onshore and contractually controlled at three points to offshore or unknown at zero. Unknown scores zero on purpose.
Points are added
No weighting. A strong contract cannot compensate for inference running offshore, and flat scoring keeps that visible.
Score becomes a percentage of 24
Seventy-five percent and above is sovereign. Forty to seventy-four percent is onshore stored, offshore processed. Below forty is offshore.
Check the answer against the contract
The quiz reflects what you believe. Confirm each answer against the vendor's terms, data processing agreement and subprocessor list before relying on the result.
The Questions Nobody Asks the Vendor
Region of storage is the question everyone asks. These are the ones that decide where the data really goes.
Region of inference
The model that reads your prompt and writes the answer runs on specific hardware in a specific place. Some vendors let you pin that region, some only pin storage, and some reserve the right to route processing anywhere. Ask where inference runs and whether that is guaranteed in writing.
- Ask for the inference region, not only the storage region
- Check whether the region is a guarantee or a preference
- Check whether failover can move processing offshore
- A private deployment fixes the region by design
Retention and training exclusion
Business and enterprise tiers from vendors such as OpenAI and Microsoft typically exclude customer content from model training and define retention, while consumer tiers often do not. The tier matters more than the brand. Read the terms for the tier your staff are on.
- Confirm training exclusion in the contract for your tier
- Confirm how long prompts and outputs are retained and why
- Check whether abuse monitoring involves human review of content
- Check what happens to data on termination
Subprocessors
The vendor may be onshore and the model provider behind it may not be. Subprocessor lists show who else touches the data and where. A tool built on a foreign model API inherits that API's residency, whatever the front-end vendor says.
- Request the current subprocessor list and its update notice process
- Trace the model provider, not only the application vendor
- Check whether you can object to a new subprocessor
- Cross-border disclosure of personal information engages APP 8 obligations
Jurisdiction
Data held by a foreign company, or by an Australian subsidiary of one, may be reachable under that country's laws regardless of where the servers are. That exposure is a legal question, not a technical one, and it is the reason some organisations require Australian ownership as well as Australian hosting.
- Identify the governing law of the contract
- Identify the ownership of the entity holding your data
- Ask your legal adviser about foreign access exposure for your sector
- On-premises or Australian-owned hosting removes most of the question
Next Steps
AI Data Sovereignty Checklist
The working checklist for keeping AI data in Australia across storage, processing and contracts.
Work through the checklist →Sovereign AI in Australia
What sovereign deployment means in practice and who needs it.
Read the guide →AI Usage Policy for Australian Organisations
Set the rules for which tools staff may use with which data.
Build the policy →Frequently Asked Questions
For some policies, yes on paper. In practice the prompt, the retrieved context and the answer all pass through wherever inference runs, and that is often not the storage region. If your requirement is that data not leave Australia, you need inference onshore as well as storage, and you need it in writing.
It varies by vendor, tier and product, and it changes. Some offer regional storage, fewer guarantee regional inference, and the position is set out in the business terms and data processing documents rather than the marketing page. Check the documents for your specific tier and keep a copy, because terms are updated.
Region of storage, region of inference, retention, training and subprocessors all become your decisions. Hosted in an Australian cloud region or on your own hardware, the data does not leave the country, nothing is used for training, retention follows your policy and the subprocessor list is short and known. Jurisdiction depends on who owns the hosting, which is also your choice.
Because a residency position you cannot evidence is not a position. If an auditor, a client or a regulator asks where the data went and the answer is that nobody checked, the practical outcome is the same as if it went offshore. Scoring unknown as zero makes the gap visible so it gets closed.
Yes, and that is often the weakest point. A personal or free-tier account is usually offshore on every question in this quiz. An AI usage policy that names approved tools and prohibits pasting organisational data into anything else closes most of the gap, and a private deployment gives staff an approved tool that is easier to use than the alternative.
No. Your answers stay on the page and are not recorded or transmitted. Refreshing the page clears them.
Sources and further reading
- Australian Privacy Principles (Office of the Australian Information Commissioner)
- The Privacy Act 1988 (Office of the Australian Information Commissioner)
- OpenAI business terms and data usage (OpenAI)
- Microsoft 365 Copilot: enterprise data protection (Microsoft)
Scored Offshore or Offshore Processed?
Tell us which tool your staff use and what data goes into it. We will map where it goes today and scope a private deployment in an Australian region or on your own hardware so every question in this quiz scores three.