Ottawa launches a public AI register: what government teams need to know
Ottawa has published its first government-wide AI register, outlining how artificial intelligence is being used or tested across federal departments and agencies. Treasury Board President Shafqat Ali announced that 43 organizations have contributed entries covering more than 400 systems at various stages - from exploration to deployment.
The register's purpose is straightforward: improve planning, reduce duplication, and help teams find proven solutions before building from scratch. It also adds visibility into whether systems are built in-house or procured from vendors, and how they're currently being used.
What the register includes
- Purpose and description of each AI system.
- Current or intended use (pilot, development, or production).
- Build source (internal vs. vendor-built).
- Ongoing updates from participating organizations.
Why this matters for your department
- Visibility: See who is solving similar problems and what tools they're using.
- Efficiency: Reuse patterns, models, and procurement artifacts instead of starting from zero.
- Risk management: Align with enterprise standards and reduce shadow AI.
- Budget discipline: Avoid duplicate spending by coordinating across programs.
Context: the broader federal AI push
The register is part of a larger effort led by Prime Minister Mark Carney to make public service delivery more efficient with AI-driven workflows. Recent actions include a translation pilot across six departments, which raised concerns about capturing nuances in French - a reminder that human review and quality controls remain essential.
The most recent budget outlines two structural moves: an Office of Digital Transformation to scale proven solutions across government, and a Shared Services Canada initiative - in partnership with National Defence and the Communications Security Establishment - to develop a made-in-Canada AI tool for internal use. Departments like Justice and Transport Canada are also identifying savings by automating repetitive tasks and applying advanced analytics.
What to do this quarter
- Inventory your AI: Catalogue pilots, vendor tools, and automated decision processes; ensure each has a clear problem statement and owner.
- Check the register first: Before funding new work, look for similar use cases and reach out to listed teams.
- Apply the Directive: Classify system impact, complete algorithmic impact assessments, and document data sources and evaluation plans. See the Directive on Automated Decision-Making from the Treasury Board Secretariat here.
- Build governance in: Add human-in-the-loop checkpoints, bilingual quality reviews, and model monitoring (bias, drift, performance).
- Strengthen procurement: Require transparency on training data, evaluation metrics, security controls, and exit options for vendor systems.
- Coordinate with SSC: Track Shared Services Canada's enterprise AI capabilities and reuse where possible (program page).
Translation pilot: what to watch
AI-assisted translation can speed up turnaround, but it can miss regional nuance and context. Departments should pair automated output with human reviewers, clear quality thresholds, and feedback loops - especially for public-facing or legally sensitive content.
Implications for data, privacy, and security
- Map datasets used for training and inference; confirm legal authority and retention schedules.
- Segment environments and enforce least-privilege access for models and prompts.
- For vendor tools, require data residency disclosures and acceptable use terms for prompts and outputs.
- Document incident response for model failures or unexpected behavior.
Budget signals and measurement
Ottawa plans to fund Statistics Canada to implement the Artificial Intelligence and Technology Measurement Program (TechStat). Expect more granular reporting on how AI is used across sectors and its impact on costs, service levels, and labour.
How to get value from the register
- Search for your use case (e.g., triaging inquiries, document summarization, case routing).
- Identify departments with similar mandates and request implementation notes and evaluation results.
- Standardize documentation: problem statement, model type, training data inputs, metrics, human oversight, rollout plan.
- Set a quarterly review cadence to add, update, and retire entries as pilots evolve.
Leadership checklist
- Assign an AI portfolio owner who reports to your executive committee.
- Tie projects to measurable outcomes: processing time, backlog reduction, accuracy, and client satisfaction.
- Fund small pilots with clear exit criteria; scale only after meeting target metrics.
- Publish internal playbooks so teams don't reinvent procurement, privacy, and testing steps.
Bottom line
The register gives public servants a clearer view of what's working, what's risky, and where to collaborate. Use it to cut duplicate work, strengthen governance, and accelerate projects that improve service delivery without sacrificing quality or compliance.
Helpful resources
- Directive on Automated Decision-Making (Treasury Board Secretariat)
- Practical AI courses by job role (Complete AI Training)
Your membership also unlocks:
