The question tends to arrive late. A hospital, a ministry or a large Canadian customer sends a security questionnaire, and somewhere around question 14 it asks where your data is stored, processed and backed up. The honest answer is often “mostly in Canada, we think.”
Picking a Canadian AWS region is the easy part. The harder part is everything around the database: backups, logs, the AI model your app calls, the email and SMS providers, the analytics script, the support desk. Any one of them can quietly send personal information to another country.
This article is for Canadian companies, public-sector suppliers and European companies with Canadian customers who need data kept in Canada. It covers what that means in practice, what has to move with the main system and how to write it down so a buyer believes it. It isn’t legal advice. Your privacy officer or counsel decides what the rule requires; our job is to build to it and show where the data actually goes.
1. Read the requirement word by word
“Data residency” gets used for several different promises. Some contracts only ask that data be stored in Canada. Others say stored and processed. A few go further and limit who may access the data from outside the country, which reaches your support staff and your vendors too.
The wording usually comes from a contract clause, a procurement checklist or a sector rule, such as a provincial health-privacy law like Ontario’s PHIPA. PIPEDA governs how private-sector organizations handle personal information more broadly. Before you design anything, get the exact sentence the buyer will hold you to, and ask what it covers: production data only, or backups, logs and copies in test environments as well.
For managers: Ask the buyer for the exact clause and confirm in writing whether it covers storage, processing, backups and access.
2. Two AWS regions in Canada
AWS has two regions in the country. Canada (Central), code ca-central-1, is in the Montreal area and has been around longer, so it carries the wider set of services. Canada West, code ca-west-1, is in Calgary. With both, you can keep a disaster-recovery copy in a second Canadian region without the data ever leaving the country.
On AWS, you choose the region resource by resource. AWS says it doesn’t move your content out of the regions you pick unless you tell it to, so the risk sits mostly in your own configuration and in the other services you connect. Not every AWS service or feature is offered in both Canadian regions, either. Check the regional service list against your design while the architecture is still on paper.
For managers: Confirm every AWS service you plan to use is offered in your Canadian region before the design is final.
3. Backups and logs are where data slips out first
A database in Montreal with snapshots copied to a US region for “safety” no longer meets a Canada-only clause. The same goes for S3 buckets replicated to another region, a backup script someone wrote years ago, or a disaster-recovery plan that names us-east-1 because it was the default in a tutorial.
Logs are sneakier. Application logs often contain names, email addresses or whole request bodies. If those logs ship to a monitoring or error-tracking service hosted in the US, personal information went with them. Keep logs in CloudWatch in the same region, or check where your logging vendor stores data, and strip personal fields before anything leaves the application. Our own default for clients with a residency requirement is to keep backups and logs in Canada alongside the main system.
For managers: List every place a backup, snapshot or log line is written, and check the region of each one.
4. AI model calls count as processing
When your app sends a document or a patient note to a language model, the model provider processes that data wherever its servers run. A call to a hosted AI API in the US is a cross-border transfer, even if your database never left Montreal.
If you use Amazon Bedrock, check which models are offered in your Canadian region, and check whether any cross-region inference setting could route requests to another country. Do the same for speech-to-text, translation and embedding services. We went through this exercise for Notermed, our AI medical scribe, which runs its AI on Bedrock under a HIPAA agreement with AWS. Its patients and agreements are in the US, so its boundary is the US. When a healthcare client needs Canadian residency, we draw the same boundary around Canada and keep the AI providers inside it.
For managers: Treat every AI provider as a place your data goes, and confirm its region before the first prompt is sent.
5. The tools around the app: email, SMS, analytics and support
A transactional email holds a name and an address, and often more. A text-message reminder holds a phone number and an appointment. Session-recording analytics can capture what someone typed into a form. A support desk or CRM keeps the conversation history. Each of these is a vendor processing personal information, and many of them run in the US by default.
Some offer a Canadian or region-pinned option, some don’t. Where you can, send email through a service in your own AWS region, keep personal data out of analytics events, and configure support tools so tickets don’t copy whole records. Where a vendor can’t keep data in Canada, the buyer should hear that from you, not discover it in an audit.
For managers: Look past the app itself: email, SMS, analytics and support tools handle personal information too.
6. Check your sub-processors one by one
Make a simple table. One row per vendor that touches personal information, with what data it receives, where it stores and processes that data, whether you have a data processing agreement, and how the vendor tells you about changes. Most established vendors publish a sub-processor list. Read it, because your email provider may in turn use a hosting company or support partner in another country.
Then draw the data flow: where data enters, every system it passes through, and where it rests. A diagram on one page catches things a table misses, like an export that a staff member downloads to a laptop every Friday.
For managers: Keep a sub-processor table and a one-page data flow diagram, and update both whenever a vendor is added.
7. How to document it for a buyer
Procurement teams want short, specific and true answers. A one- or two-page residency statement usually covers it: the AWS region or regions, where backups and logs live, the sub-processor table with locations, encryption at rest and in transit, who has administrative access and from where, how you’d notice a breach, and how a change to any of this is approved.
Back the words with controls a reviewer can check. With infrastructure in Terraform, the region is written in code and reviewed like any other change. AWS Organizations lets you set service control policies that deny actions outside approved regions, so a stray resource in another country fails at creation instead of turning up months later. For European companies serving Canadian customers, the same setup lets you run EU regions for EU data and ca-central-1 for Canadian data, with a statement for each.
For managers: Write a residency statement you’d be comfortable defending line by line, and back it with controls rather than intentions.
The short version
Keeping data in Canada starts with the region and doesn’t end there. Backups, logs, AI calls, email, SMS, analytics and support tools all count. Get the exact requirement in writing, map every place personal information goes, check each vendor and document the result in plain language. If you can show a buyer that map, the questionnaire becomes the easy part.