Canada's New AI Registry: What Government Teams Need to Know and Do Next
Classifying insects on farms. Scanning packages to spot guns. Plotting safe medical treatment for astronauts. These are not hypotheticals-they're live or exploratory use cases inside the Government of Canada's new public AI registry, which lists 400+ systems across 42 departments.
The registry is now live on Canada's Open Government website. You can search by department, use case, and vendor, giving public servants a clearer view of how AI is being applied across the federal service. View the registry.
What the registry covers
- Only "notable" AI projects are listed. Low-risk utilities like spell-check and virtual assistants are excluded.
- Each entry includes the department, the purpose of the system, and the vendor (if applicable).
- This is an early-stage release with basic search; more detail and features are planned based on public feedback next year.
Examples already in use or testing
- Package screening for firearms: the Canada Border Services Agency (CBSA) is testing a tool from Israeli firm SeeTrue.
- Agriculture: AI labeling to classify insects found on farms and in food contexts.
- Space health: tools that help calculate treatment plans for astronauts.
- Longest-running use: since 1994, CBSA has used "fuzzy search" to flag high-risk travellers by searching names across systems.
Transparency-and why it matters inside government
"Artificial intelligence is transforming governments, and we are committed to providing Canadians with information about how it is being used to support programs and services. The AI Register is an important step in building public trust and ensuring the responsible use of AI across the federal public service," said Treasury Board president Shafqat Ali.
Worth noting: the registry page itself was created using machine translation, with human review of outputs.
Vendors, concentration risk, and procurement opportunity
At first glance, many listed systems look in-house and fairly straightforward. That said, without details on architecture, training data, and testing, it's tough to judge adherence to best practices.
Vendor concentration stands out. Microsoft appears 75 times and OpenAI 15 times in the document. Arvind Gupta, a University of Toronto computer science professor and member of Ottawa's AI Strategy Task Force, suggested the registry could be a springboard for a more deliberate procurement approach that creates room for Canadian SMEs.
He pointed to the US Small Business Innovation Research (SBIR) model, which asks departments to set aside 2% of discretionary budgets for novel solutions from smaller firms (the program is currently paused pending Congressional funding). Learn more about SBIR here: sbir.gov.
Practical steps for departments and agencies
- Inventory your AI: Confirm that your internal and vendor-supported systems are accurately represented in the registry. Close gaps and retire stale entries.
- Clarify risk tiers: Distinguish low-risk utilities from higher-impact systems with real consequences for people, dollars, or decisions.
- Check data provenance: Document training data sources, data-sharing agreements, and any sensitive attributes. If you can't explain the data, you can't defend the output.
- Strengthen testing: Establish clear performance thresholds, bias testing, drift monitoring, and fallback plans. Log decisions where AI influences outcomes.
- Broaden the vendor bench: Avoid reliance on a small set of large providers. Pilot SMEs where feasible, especially for well-scoped, high-value problems.
- Sharpen procurement: Use modular contracts, clear evaluation rubrics, and phased milestones. Ask vendors for model cards, evaluation reports, and audit hooks up front.
- Upskill your team: Build AI literacy across policy, procurement, privacy, and IT so decisions are consistent and defensible. For structured options by role, see this resource: AI courses by job function.
- Engage early with oversight: Pull in legal, security, and privacy teams before pilots go live. Document approvals and accountability lines.
What's coming next
The government describes this release as an early version. Expect expanded fields, better search, and more frequent updates next year, shaped by public and internal feedback. Departments should treat the registry as both a transparency tool and a management tool-one that pressures all of us to clean up inventories, reduce risk, and get better at delivering value with AI.
Image courtesy Unsplash. Photo by Nick Linnen.
Your membership also unlocks:
