← Back to Insights
Canadian AI

AWS Canada (Central) and the Limits of True Data Residency

An AWS server in Montreal doesn't mean your data stays Canadian. Here's what 'Canada (Central)' actually promises and what real Canadian AI looks like.

By Augure·
The Canadian flag waving on a pole before the Three Sisters mountains in Canmore

Short version: no, not by itself. AWS has a data centre region called Canada (Central), based in Montreal, and storing your data there keeps it physically in the country. But AWS is a US company, bound by US law, and that one fact changes what "in Canada" can actually guarantee you.

This matters more than it used to. Canadian businesses are buying AI tools faster than most can read the privacy policy behind them. A lot of those tools point proudly to a "Canada" region and call it a day. That's useful. It's also not the whole story.

What does AWS Canada (Central) actually store in Canada?

Amazon opened the Canada (Central) region in Montreal in December 2016. If you pick that region when you configure a service, your data sits on hardware physically located in Quebec. That part is real, and it satisfies the "store data in Canada" requirement that shows up in a lot of provincial privacy rules and government procurement contracts.

What it doesn't do is change who runs the servers. Amazon Web Services is a subsidiary of Amazon.com, Inc., a US company headquartered in Seattle. The region is Canadian. The company operating it is American.

Why does the company's home country matter if the servers are here?

Because of the CLOUD Act, a 2018 US law. It lets US courts order US companies to hand over data those companies control, no matter where the servers physically sit. A US court order aimed at Amazon can, in principle, reach data stored in Montreal, because the order targets the company, not the country.

This doesn't mean AWS Canada (Central) is insecure, or that Amazon hands data over to anyone who asks. Amazon fights a lot of these requests and publishes transparency reports about it. But the legal pathway exists, and it exists because of corporate ownership, not server address. A Canadian data centre run by a US parent company is still, legally, a US company's data centre with a Canadian postal code.

A skeptic might point out that the CLOUD Act requires a US court order, and those aren't handed out casually. That's fair. The order has to come from a US judge, usually tied to a criminal investigation or a warrant request, and Amazon can and does push back in court before complying. But the process still bypasses any Canadian court entirely. No Canadian judge reviews the request, no Canadian law gets a say, and your business has no seat at that table. The existence of friction in the process isn't the same as the process not applying to you.

Quebec's privacy law asks organizations to assess risk before sending personal data outside the province, including the legal protections that apply wherever it lands. If the company holding your data can be compelled by a foreign court no matter where the hardware sits, that's exactly what the assessment is supposed to catch.

So what counts as real Canadian AI?

This is where the distinction actually bites. A genuinely Canadian AI platform isn't just a company that checked the "Canada region" box in a cloud console. It's a company incorporated in Canada, owned by Canadians, with no US parent and no US investors sitting above it in the org chart.

That structural difference is what closes the CLOUD Act pathway, not the data centre's postal code. If a company has no US corporate parent and no US investors, there's no US entity for a US court to compel over that company's customer content.

Augure is built this way. Augure has no US parent company and no US investors, and customer data is stored in Canada because Canadian privacy law requires it. Inference, the actual AI processing, runs on Canadian infrastructure, with some model tiers served by vetted EU partners under zero-data-retention agreements for reliability. Customer conversations and documents are never routed to a US provider. Customer data is never used to train models. Where the data sits and who's legally allowed to ask for it are two different questions, and it's worth keeping them separate.

To give the full picture: a small amount of processing still touches US infrastructure, for things like payment card networks and email delivery. That's disclosed plainly rather than buried, because the real comparison isn't zero US contact anywhere. It's whether a US court has a legal pathway to your customer content. For Augure's customer content, it doesn't.

What does this cost me, practically?

Here's the part that actually affects a nine-person business. If you're using a US-parented AI tool today, even one with a Canadian region option, your Law 25 risk assessment should account for US legal reach over that provider. That's extra diligence work, and it's ongoing, not a one-time checkbox.

That diligence has a shape, not just a feeling. It means tracking which vendor owns which tool, re-checking that ownership when a vendor gets acquired, and documenting the legal basis for every cross-border flow you rely on. For a nine-person shop with no compliance staff, that's usually an afternoon every quarter, assuming nothing changes. If a vendor gets bought by a US company mid-contract, which happens more often than you'd think in this industry, the whole assessment has to be redone on short notice.

Switching to a Canadian-owned AI platform removes that line item from your risk assessment. It doesn't erase every compliance task. You still need to know what data you're putting in, who on your team can see it, and how long you keep it. No vendor removes that responsibility from you. But it does remove one specific, recurring question from every vendor review you'll ever run again.

A Canadian region on a US company's cloud is still a US company's cloud.

Is this worth switching tools over?

Depends on what you're putting into the tool. If your nine-person shop is using AI chat to brainstorm blog titles, the CLOUD Act is not your biggest risk today. If you're uploading client files, contracts, or anything with personal health or financial information in it, the ownership question stops being theoretical.

Augure starts free — a daily compute allowance, ten documents, basic web search, no credit card. Paid plans start at $20 a month for unlimited messages and persistent memory, where the AI remembers your past conversations instead of starting cold every time. For law firms reviewing contracts and NDAs, Augure Legal runs compliance checks for Law 25 and PIPEDA, the federal privacy law, starting at $149 a month for a solo practitioner.

None of this guarantees you're compliant. No software can promise that on your behalf. What it does is remove one entire category of risk from the list you have to manage yourself.

What to do this week

  1. Pull up the privacy policy of every AI tool your team uses and find the parent company. If it's a US entity, note it.
  2. Check what you've actually uploaded to those tools in the last 90 days — client files, contracts, anything personal or financial.
  3. Try Augure's free plan for one real task this week, the kind you'd normally run through a US tool, and compare.

Start at augureai.ca.

Where this comes from: The CLOUD Act is the US Clarifying Lawful Overseas Use of Data Act, passed in 2018. Quebec's privacy law provisions on cross-border transfer risk assessment appear in what's commonly called Law 25, section 17.

A

About Augure

Augure is a sovereign AI platform for regulated Canadian organizations. Chat, knowledge base, and compliance tools — all running on Canadian infrastructure.

Ready to try sovereign AI?

Start free. No credit card required.

Get Started