Treasury Board Directive on Automated Decision-Making (with Algorithmic Impact Assessment)
Last reviewed Sep 25, 2026.
In short
The Directive governs how federal departments use automated systems to make or support administrative decisions about people. Before launching a system, a department must complete and publish an Algorithmic Impact Assessment that scores the system into one of four impact levels; higher levels trigger peer review, human-in-the-loop decision-making, more testing and monitoring. Plain-language notices, explanations of decisions, bias testing and data-governance duties apply. The 2023 amendments extended it to internal services such as hiring and required the AIA to be published before launch.
Who it applies to
- Organisations that build or use AI systems in Canada (federal).
- Use cases in scope: public sector.
- Size: everyone, including solo operators.
- Binding on federal government departments; relevant to vendors selling automated decision systems to the Government of Canada and a useful template for provinces and municipalities.
Key dates
- Apr 1, 2019✓Directive took effect for federal institutions
- Apr 25, 2023✓Third-review amendments in force (scope extended to internal services such as hiring; AIA published before launch)
- Apr 25, 2024✓Existing systems had to comply with the amended Directive
What you have to do
- Complete the Algorithmic Impact Assessment before development or procurement and publish it on the Open Government portal before launch.
- Apply the safeguards for the resulting impact level: peer review, human intervention, training and contingency plans.
- Give people notice that an automated system is used and a meaningful explanation of decisions.
- Test data and outputs for bias before and after launch; validate and monitor regularly.
- If you sell to government, be ready to provide source-code access, documentation and testing evidence required by the Directive.
Penalties
Administrative: non-compliance is handled under the Treasury Board's compliance framework (Framework for the Management of Compliance), not by fines. For vendors, it becomes a contract condition.
Related lessons
- Level 1The world map of AI rules: EU, US patchwork, Canada, standards8 min
- Level 2Build your AI inventory6 min
- Level 2Risk classification: prohibited, high-risk, limited, minimal8 min
- Level 3Human oversight that actually works6 min
- Level 3Documenting decisions: records and impact assessments7 min
Industries where it matters
Sources
- Directive on Automated Decision-Making — Treasury Board of Canada Secretariat ↗
- Amendments to the Directive on Automated Decision-Making — Canada.ca ↗
- Algorithmic Impact Assessment tool — Open Government ↗
Last reviewed Sep 25, 2026.