AI regulation in Canada after AIDA was withdrawn
Canada: AIDA (withdrawn), Directive on Automated Decision-Making, Law 25 (Quebec)
- Status
- Withdrawn
- Binding
- No
- Object analysed
- Automated decision
- Requirements
- 3
- Next milestone
- 2027
In short
The federal AIDA died on the order paper in January 2025. What remains: the federal Directive (public sector, 4-level Algorithmic Impact Assessment) and Quebec Law 25 on automated decisions.
Steps to compliance
- Qualify each AI systemAxes to decide: Impact level (DADM).
- Determine your roleDuties vary by role: Federal institution and Enterprise.
- Apply the 3 requirementsThey focus on: Impact assessments, Accuracy & robustness, Fairness & non-discrimination and Human oversight.
- Prove it with checks6 checks to document, 6 of which also serve NIST AI RMF, Colorado AI Act and US federal.
- Track the deadlinesNext milestone: 2027, New federal AI bill (potential).
Scope and penalties
- Kind
- Bill
- Scope
- Federal institutions; businesses processing Quebec residents' data.
- Territorial reach
- Canada, Quebec.
- Penalties
- Law 25: up to CAD 25M or 4% of worldwide turnover.
- Jurisdiction
- Canada
Timeline
Qualifying a system
Classification axes and possible verdicts
Impact level (DADM)
Requirements
3 requirements
| Code | Article | Requirement | Applies to | Checks |
|---|---|---|---|---|
| CAN-01 | DADM §6.1 | Algorithmic Impact Assessment (AIA) before production | ||
| CAN-02 | DADM §6.3 | Quality assurance, bias testing and human intervention by level | ||
| CAN-03 | Loi 25, art. 12.1 | Inform of a fully automated decision and allow submissions to a human |
Checks to document
Evidence collected for a check counts for every regulation that uses it.
| Code | Check | Scope | Themes | Also used by |
|---|---|---|---|---|
| CHK-IMPACT-ASSESS | An impact assessment is performed, documented and used in go/no-go and risk decisions | System | ||
| CHK-TEVV | TEVV plan, test sets, metrics and data considerations are documented | System | ||
| CHK-BIAS | Fairness and bias are evaluated and results documented | System | ||
| VER-008-02 | System designed to allow human oversight (stop button, override) | System | ||
| VER-029-D-01 | Affected persons informed of the use of the AI system | System | ||
| VER-033-D-01 | Decision explanation procedure documented | Organisation |
Themes covered
Frequently asked questions
Who is in scope of Canada?
Federal institutions; businesses processing Quebec residents' data. Canada, Quebec.
What penalties does Canada carry?
Law 25: up to CAD 25M or 4% of worldwide turnover.
When do the Canada obligations apply?
Sep 27, 2023: Voluntary code of conduct on generative AI; Jan 6, 2025: Prorogation: AIDA dies on the order paper; 2027: New federal AI bill.
Is Canada binding?
No. Kind: bill. Status: withdrawn.
How does Canada relate to other regulations?
The same checks serve several texts. Shared checks: NIST AI RMF (4), Colorado AI Act (4) and US federal (3).
Related regulations
Official sources
Data checked on Sep 25, 2026. General information, not legal advice. Check the official texts and get advice for your situation.
Run these requirements across all your AI systems
TrustFlow inventories your systems, qualifies them under each regulation and collects evidence once for every referential.