Bureaucratic Silences: What the Canadian AI Register Reveals, Omits, and Obscures
Summary: arXiv:2604.15514v1 Announce Type: new
In November 2025, the Government of Canada operationalized its commitment to transparency by releasing its first Federal AI Register. This initiative aimed to provide a comprehensive overview of the AI systems deployed within the public sector. However, this article argues that such registers are not neutral mirrors of government activity, but rather active instruments of ontological design that configure the boundaries of accountability.
Introduction
The Canadian Federal AI Register was introduced as a step towards greater governmental transparency. It was expected to serve as a detailed account of the AI systems utilized in various public sector contexts. Nevertheless, our analysis reveals that the Register’s construction and presentation of information reflect underlying biases and omissions that merit critical examination.
Methodology
To investigate the implications of the Federal AI Register, we employed the Algorithmic Decision-Making Adapted for the Public Sector (ADMAPS) framework. This methodology allowed us to combine quantitative mapping of the Register’s complete dataset of 409 systems with deductive qualitative coding, providing a comprehensive understanding of the data presented.
Findings
The analysis uncovered several key findings that highlight the discrepancies between the proclaimed goals of the AI Register and the actual bureaucratic practices observed:
- Rhetoric vs. Reality: There exists a sharp divergence between the rhetoric of “sovereign AI” and the reality of bureaucratic practice. While 86% of systems are deployed internally for operational efficiency, the Register fails to adequately represent the complexities involved in their deployment.
- Obscured Human Discretion: The Register systematically obscures the critical human discretion, training, and uncertainty management essential for the effective operation of AI systems.
- Technical Descriptions Over Context: By prioritizing technical descriptions, the Register constructs an ontology of AI as “reliable tooling,” thereby neglecting the sociotechnical context that influences decision-making processes.
Implications for Accountability
The implications of our findings are significant. The current design of the Federal AI Register risks transforming transparency into a mere compliance exercise, where visibility does not equate to accountability. The lack of contestability in the presented information undermines the potential for public scrutiny and engagement, which are crucial for maintaining democratic oversight of AI systems in the public sector.
Conclusion
In conclusion, while the Canadian Federal AI Register represents an important step towards transparency, it is crucial to recognize its limitations. Without a critical reassessment of its design and presentation, such transparency artifacts may inadvertently automate accountability into a performative exercise. To truly foster accountability and public trust in AI systems, a shift in the conceptualization and design of such registers is necessary.
As we move forward, it is essential that policy-makers, researchers, and the public remain vigilant in scrutinizing the narratives constructed by government registers and advocate for more inclusive and transparent frameworks that reflect the complexities of AI decision-making.