A University of Toronto study of all 409 systems in Canada's federal AI register found uneven disclosure and heavy third-party dependence in some agencies, while 86% of listed systems served internal operations. Ottawa Citizen reported on Aug. 14 that three of four CRTC systems were developed by Microsoft and that the register had grown to 412 entries.
A University of Toronto research team analyzed all 409 systems in the Canadian federal AI register and found that the inventory reveals broad adoption while omitting important context about human discretion, training and accountability. The paper, published for the 2026 ACM Conference on Fairness, Accountability, and Transparency, used quantitative mapping and qualitative coding based on the ADMAPS public-sector framework.
The researchers reported that 86% of registered systems were intended for internal operations. They also found that 44% were in development and 39% were already in production, suggesting agencies are expanding and maintaining AI systems at the same time.
Vendor dependence varies by agency
Ottawa Citizen reported on Aug. 14 that some agencies used only internally developed tools while others relied heavily on external vendors. Lead author Dipto Das told the newspaper that Microsoft, OpenAI and Google appeared among those third parties. The report said three of the four systems listed for the Canadian Radio-television and Telecommunications Commission were developed by Microsoft.
That finding does not mean every federal agency has the same dependency profile. The paper and newspaper both emphasize uneven reporting across institutions, and the public register covers 42 organizations even though the federal government has more than 200 departments and agencies. Ottawa Citizen said the register had grown from the study's 409-system snapshot to 412 entries.
Visibility is not the same as accountability
The paper's central argument is that a register can disclose systems while still obscuring who exercises judgment, what training or staffing is required, and how uncertainty is handled. The authors warn that technical descriptions may make AI look like reliable tooling even when outcomes depend on organizational choices and human review.
For public-sector data and ML teams, the practical gap is documentation depth. A useful inventory should identify the responsible organization, vendor and hosting dependencies, lifecycle stage, data categories, human decision points, validation process and a route for contesting outcomes. Counts alone establish adoption breadth; they do not establish whether a system is safe, effective or accountable. Evidence for Democracy's separate March review provides background on the same federal register, including more than 400 recorded uses across 42 agencies, but it is not the University of Toronto study that prompted the August report.
Key Points #
- 1The University of Toronto team analyzed the register's complete 409-system snapshot using quantitative mapping and qualitative coding.
- 2The study found 86% of systems served internal operations, with 44% in development and 39% in production.
- 3Vendor dependence varied by agency; Ottawa Citizen reported that Microsoft developed three of four CRTC systems in the register.
- 4The authors argue that system inventories can provide visibility while still obscuring human discretion and accountability.
Scoring Rationale #
The study provides a complete-register analysis with reproducible counts and concrete governance implications for public-sector AI inventories. The register's uneven disclosure limits conclusions about system performance or agency-wide risk.
Sources #
Primary source and supporting public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
[Active Search Campaigns by BudgetEasy](/problems/sql/active-search-campaigns-by-budget)
[High CPC Clicks & Poor Landing PagesMedium](/problems/sql/high-cpc-clicks-poor-landing-page)
[Campaign ROAS by Attribution ModelHard](/problems/sql/campaign-roas-by-attribution-model)
250 free problems · No credit card