Directive on Automated Decision-Making

Record CA-01

Issued by: Treasury Board of Canada Secretariat
Jurisdiction: Canada
Level · region: National · North America
Published / adopted: 2019-03-04
In force: 2019-04-01
Last updated: 2025-06-24
Language: English, French

Summary

Mandatory Treasury Board directive for federal institutions that use automated decision systems, including machine learning, to make or assist administrative decisions about clients. Before production, institutions must complete and publish an Algorithmic Impact Assessment that sets one of four impact levels and apply that level's requirements: notice and explanation, peer review, Gender-based Analysis Plus, staff training, testing and monitoring for bias and unfair outcomes, human involvement (a human makes the final decision at levels III and IV), approval up to Treasury Board, recourse and public reporting.

Safety topics

public-sector userisk-based classificationhuman oversighttransparency & labellingevaluation & testingaccountability & liability

Status notes

Binding on the institutions subject to the Policy on Service and Digital, under whose authority it is issued; other departments and agencies are encouraged to meet it as good practice. Launched on 4 March 2019, it took effect on 1 April 2019, with compliance required by 1 April 2020, and applies to systems developed or procured after that date. Amended on 1 April 2021, on 25 April 2023 (third review: expanded scope and new measures on explanation, bias testing, data governance, GBA Plus and peer review) and on 24 June 2025 (fourth review: transparency of systems and their impacts, senior-level accountability, testing and monitoring; the Algorithmic Impact Assessment tool was updated with it). Systems developed or procured before 24 June 2025, and agents of Parliament, had until 24 June 2026 to comply with the new or updated requirements. The directive is reviewed every two years.

Sources

Open the official source https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=32592

Secondary source: https://www.canada.ca/en/government/system/digital-government/digital-government-innovations/responsible-use-ai/progress.html

Local copy

Archived. Last attempt 2026-10-17 22:11:34 UTC

Official source (third-party copy, not published) archive/CA-01/2026-10-11/official.html

Retrieved 2026-10-11 13:59:29 UTC

SHA-256 225f990673205a49cee216fb8fc124909e8eb917559e3e258a8dd7fb7c35a0e3

Secondary source (third-party copy, not published) archive/CA-01/2026-10-11/secondary.html

Retrieved 2026-10-11 13:59:31 UTC

SHA-256 b5233ca898db5f5b66e90dd936b55aba36b88d45e1dec7ea93c44eff08ecde69

Verification

Another official source, such as an announcement by the issuer, confirms the title, date and status. checked 2026-10-18 · alternative-official-url-fetched

Recheck of 17 Oct 2026: re-read the directive's page on tbs-sct.canada.ca (stored 11 Oct 2026; the archiver found it unchanged on 17 Oct 2026) and TBS's 'Progress on AI in government' page on canada.ca (stored the same day as the secondary source, also unchanged). The directive page confirms the title, the effective date of 1 April 2019 with compliance by 1 April 2020, 'Date modified: 2025-06-24', the transition to 24 June 2026 for existing systems and agents of Parliament, and the requirements and impact levels; its archive block lists the 2021-03-24 and 2023-04-24 versions as replaced. The launch date of 4 March 2019 and the amendment dates of 1 April 2021 and 25 April 2023 rest on the progress page, which is stored as the secondary source, so date_published is supported by another official source, not by the directive page itself.

Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 7 field(s) supported.

History

2026-10-18related: [{"slug": "ca-ai-strategy-federal-public-service-2025-2027", "id": "CA-06"}, {"s → [{"slug": "ca-ai-strategy-federal-public-service-2025-2027", "id": "CA-06"}, {"s — two-way link added: CA-ON-01 link(s) to CA-012026-10-18-fix-related-links · Claude Code (claude-opus-5-5)
2026-10-18verification.level: A → B — level computed from the evidence by tools/pipeline.py2026-10-17-recheck-north-america-1 · Claude Code (claude-opus-5-5)
2026-10-18notes: Binding on federal institutions as Treasury Board policy (not a statute). Launch → Binding on the institutions subject to the Policy on Service and Digital, under — the transition to 24 June 2026 for systems procured before 24 June 2025 and for agents of Parliament (section 1.2) added, with its outcome; application section 8 and the content of each review added; 'effective April 2019' made exact (1 April 2019)2026-10-17-recheck-north-america-1 · Claude Code (claude-opus-5-5)
2026-10-18summary: Mandatory Treasury Board policy instrument requiring federal institutions that u → Mandatory Treasury Board directive for federal institutions that use automated d — rewritten from the directive text to show its structure: the Algorithmic Impact Assessment sets one of four impact levels (Appendix B) and the requirements follow the level (Appendix C), including a human final decision at levels III and IV and approval up to Treasury Board2026-10-17-recheck-north-america-1 · Claude Code (claude-opus-5-5)

Related records

GC AI Strategy 2025-2027 Canada · Published (non-binding, current)

GC Generative AI Guide Canada · Published (non-binding, current)

Ontario Bill 194 / Enhancing Digital Security and Trust Act Ontario (Canada) · In force

Citation

Treasury Board of Canada Secretariat. Directive on Automated Decision-Making. 2019-03-04. https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=32592. Status: In force; verified 2026-10-18 (level B). Local copy archive/CA-01/2026-10-11/official.html retrieved 2026-10-11 13:59:29 UTC, SHA-256 225f990673205a49cee216fb8fc124909e8eb917559e3e258a8dd7fb7c35a0e3. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record CA-01.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.