What Canadian municipal buyers now ask about AI and where their residents' data goes. cover

What Canadian municipal buyers now ask about AI and where their residents’ data goes.

Canadian municipal AI procurement now asks more than whether data is stored in Canada. Procurement teams want to know which AI features are switched on, what the model does with a resident's request, which countries the data passes through, and what paperwork the city must complete before any of it goes live.

That shift comes partly from the arrival of AI in ordinary municipal software, from permit portals to call centre tools, and partly from changes in provincial privacy law. This article sets out the questions we now see municipal buyers ask, and the legal rules behind them in British Columbia and Quebec, with the federal Directive on Automated Decision-Making as a useful model.

For the wider picture of public sector buying, start with our guide to Canadian public sector AI procurement.

Canadian municipal AI procurement: beyond residency

Residency still matters, but it answers only part of the question. An AI feature can keep stored records in a Canadian data centre while sending each prompt to a model hosted somewhere else. Support staff in another country may be able to see logs. A subprocessor may keep copies for abuse monitoring. None of that shows up in a one line residency promise.

So buyers now ask about the whole route the data takes: where it is collected, where it is processed, where it is stored, who can access it and for how long. That is a better question, and in some provinces it is the one the law now asks.

Canadian municipal AI procurement traces resident data through cloud, logs and support systems.

British Columbia: from a residency rule to an assessment

British Columbia used to have one of the strictest rules in the country. Its Freedom of Information and Protection of Privacy Act required personal information held by public bodies, including municipalities, to be stored and accessed in Canada except in limited cases.

That changed with the Freedom of Information and Protection of Privacy Amendment Act, 2021, which received royal assent on 25 November 2021. Section 33.1 now says a public body may disclose personal information outside Canada only if the disclosure is in accordance with the regulations made by the minister. The Personal Information Disclosure for Storage Outside of Canada Regulation requires an assessment of any program, project or system in which sensitive personal information is disclosed to be stored outside Canada.

That assessment sits inside the privacy impact assessment. The provincial guidance says it should identify the privacy risks of storing the information outside Canada and the responses to them, so the head of the public body can make an informed decision. The guidance lists health, financial, biometric and location data among the kinds of information commonly treated as sensitive, and it stresses that context matters.

For a B.C. municipality, data location feeds a legal assessment. Canadian municipal AI procurement needs clear vendor answers.

Quebec: an assessment and a written agreement

Quebec took a different route through Law 25, the Act to modernize legislative provisions as regards the protection of personal information. For public bodies, which include municipalities, the relevant rule is section 70.1 of the Act respecting Access to documents held by public bodies and the Protection of personal information. The current version has been in force since 22 September 2023.

Section 70.1 says that before releasing personal information outside Quebec, a public body must conduct a privacy impact assessment. The assessment must consider the sensitivity of the information, the purposes for which it will be used, the protection measures, including contractual ones, and the legal framework in the place where it will go. The information may be released only if the assessment shows it would receive adequate protection, and the release must be the subject of a written agreement that reflects the assessment.

The same rule applies when the public body asks a person or body outside Quebec to collect, use, release or keep information on its behalf. Note that the line is the Quebec border. A vendor hosting in Ontario is outside Quebec for this purpose.

Privacy assessment and written agreement protect resident data across provincial boundaries.

The federal model buyers borrow from

The Government of Canada's Directive on Automated Decision-Making applies to federal institutions. Municipalities are outside its scope. Even so, many municipal teams use it as a template because it is clear and public.

The directive took effect on 1 April 2019, and it applies to automated decision systems developed or procured after 1 April 2020. Its latest version is dated 24 June 2025. Systems developed or procured before that date had until 24 June 2026 to meet the new or updated requirements. Its companion, the Algorithmic Impact Assessment tool, scores a system's risk and sets the level of safeguards, such as peer review, notice to clients and human involvement in decisions.

For a municipality, asking a vendor to answer the impact assessment questions is a quick way to see how much it understands about its own product.

Canadian municipal AI procurement: the questions buyers now ask

Put together, these rules produce a fairly consistent set of questions. We see variations of these in tenders and vendor questionnaires.

Which AI features are in the product, and which are on by default? Buyers want a written list. They also want the ability to turn features off and confirmation that new ones will not be switched on without notice.

Where does each kind of data go? That means stored records, prompts and responses, logs, backups and support access, each with a location. A diagram helps.

Which subprocessors are involved, and where are they? A list with locations, and a commitment to give notice before it changes.

Is our data used to train or improve models? Buyers increasingly expect a contractual no, because a setting can change.

How long is each copy kept, and can we delete it? Retention for prompts and logs is often longer than people assume.

Does the AI make or shape decisions about residents? Drafting a reply to an email carries less risk than scoring a benefit application or ranking permit requests. Where the tool influences a decision, buyers ask how a person reviews it and how a resident can challenge it.

What evidence can you give us for our privacy impact assessment? In B.C. and Quebec this is the question the others serve. The city must complete the assessment, so the vendor's job is to supply the facts it needs.

Our article on how CISOs evaluate AI vendors covers the security side of the same questionnaire in more detail.

What good answers look like

The strongest vendor responses share a few traits. They name specific regions and services. They separate the core product from the AI features, because the data route is often different. They say plainly what is logged and for how long. And they offer contract terms that match their claims.

Weak responses rely on certificates alone, avoid the question of where model processing happens, or promise residency for stored data while staying silent about prompts. A buyer who gets one of those answers should ask again, in writing.

An evidence register tracks AI data routes, retention and recurring reviews.

Canadian municipal AI procurement: keeping the evidence

Canadian municipal AI procurement does not end at signature. The assessment needs revisiting. Vendors add features, change subprocessors and move workloads. Each change can alter the data route that the assessment relied on.

A simple register helps: for each AI feature in use, record what it does, where its data goes, which assessment covers it and when that was last reviewed. Our guide to building an AI evidence register shows one way to set that up, and our piece on what regulated organisations must log covers the records worth keeping.

If you want templates for vendor questions and assessments, our free AI governance handbook has them ready to adapt.

Reviewed through 6 November 2026 against B.C.'s Freedom of Information and Protection of Privacy Act section 33.1 on BC Laws, the 2021 amendment act, the Province of British Columbia's guidance on disclosures outside Canada, section 70.1 of Quebec's Access Act on Légis Québec with its version history, and the Treasury Board of Canada Secretariat's Directive on Automated Decision-Making. This article is general information, not legal advice.

A Canadian address is only one part of an AI supplier review.

Source: Directive on Automated Decision-Making.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *