社会ガイド

Canada’s Directive on Automated Decision-Making

The Treasury Board’s Directive on Automated Decision-Making is a mandatory policy for covered federal institutions using production systems to make or support administrative decisions and related assessments.

  • 3 分で読めます
  • 最終更新日
このページでは3 分で読めます
  1. 概要
  2. ディープダイブ
  3. 戦略的影響
  4. The Future of Canada’s Directive on Automated Decision-Making
  5. 現実世界の実装
  6. リスクとガードレール
  7. 実装ロードマップ
  8. 探検を続けましょう
  9. よくある質問

概要

Its Algorithmic Impact Assessment assigns an impact level that determines proportionate safeguards.

ディープダイブ

Canada’s Directive on Automated Decision-Making is a Treasury Board policy instrument, not a general statute for every Canadian employer or private-sector AI system. It applies to institutions subject to the Policy on Service and Digital when a production automated decision system makes an administrative decision or a related assessment about a client. The amended scope also includes internal services such as federal hiring, performance management, and employee security screening. Research-only systems and test environments are excluded. Other federal organizations are encouraged to follow it as good practice where it does not apply. Before production, the responsible senior official must complete, approve, and publish an Algorithmic Impact Assessment (AIA), then apply the requirements corresponding to the assessed impact level. The AIA asks about the decision, system design, data, privacy, security, transparency, recourse, and potential impacts, including accessibility and disability-related concerns. The four levels scale safeguards: higher-impact systems face more demanding peer review, notice, explanation, human involvement, testing, monitoring, and contingency requirements. The AIA must be reviewed and updated on schedule and when system functionality or scope changes. The 2025 directive update applies to new or procured systems from 24 June 2025. The transition deadline for systems developed or procured before that date was 24 June 2026 and has now passed; covered systems are expected to meet the new or updated requirements. Agents of Parliament have a transition period to the same date with specified publication exceptions. Required peer-review findings must be published before launch for applicable levels. AIA level is a risk classification within the federal directive, not a claim that a system is safe in every context. Departments should identify the decision and legal authority, assign a senior owner, complete the AIA early, and document how each triggered requirement is met. If a system is only advisory but influences an administrative outcome, the directive may still apply. A vendor’s assessment does not replace the institution’s own AIA or accountability.

戦略的影響

リスクと安全性

AI による壊滅的な被害も日常的な被害も、誰がリスクを理解し、誰が行動できるかにかかっています。

より明確な判決

国民と専門家のリテラシーは、強力な安全政策が政治的に可能かどうかを左右します。

誇大広告を打ち破る

明確な説明は、誇大広告、研究室の PR、曖昧な倫理劇場に囚われることを減らします。

The Future of Canada’s Directive on Automated Decision-Making

The directive is reviewed every two years and can receive off-cycle review. Its 2025 amendments expanded coverage and strengthened transparency, bias testing, data governance, explanations, and inclusion. Departments should check the current directive and AIA tool before procurement or launch, especially because existing systems’ transition date was 24 June 2026. Updates can alter required safeguards and transition rules. The directive is reviewed every two years and can change outside that cycle. Check updated Treasury Board policy and transition dates before launching or materially changing a federal system.

現実世界の実装

A federal department completes and publishes an AIA before launching a benefits eligibility system.

A department assesses a hiring-screening tool as an internal service under the amended directive.

A program revises its AIA after changing the model’s scope or data source.

A product team confirms a research sandbox is excluded before moving the system into production.

リスクとガードレール

  • 能力が複雑になる一方で、実存的なリスクを SF として扱います。

  • 高度な自律性の下での調整による表面製品の安全性を混乱させる。

  • 英語以外や専門家ではない聴衆には、低品質の情報源しか提供されません。

実装ロードマップ

  1. 製品の危害、誤使用、制御不能/調整不良のリスクを分離します。

  2. どのような証拠がタイムラインと重大度についてのあなたの見方を変えるかを尋ねてください。

  3. マーケティング上の主張よりも、一次情報源と具体的な評価を優先します。

  4. 意識だけでなく、キャリア、政策、資金、スキルなど、行動経路を 1 つ特定します。

探検を続けましょう

Free newsletter

Get the daily AI briefing

Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.

One email each weekday. Unsubscribe in one click. We never sell or share your address.

Test yourself

Take the Canada’s Directive on Automated Decision-Making quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

クイズを開始する

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

よくある質問

What is Canada’s Directive on Automated Decision-Making?

The Treasury Board’s Directive on Automated Decision-Making is a mandatory policy for covered federal institutions using production systems to make or support administrative decisions and related assessments. Its Algorithmic Impact Assessment assigns an impact level that determines proportionate safeguards.

What kind of organization is directly subject to the Directive?

The directive applies to institutions subject to the federal policy.

When must the AIA be completed and published?

The directive requires approval and publication before production.

What does the Algorithmic Impact Assessment determine?

AIA level drives proportional requirements under the directive.

Did the 2025 amendments expand the directive to internal federal services?

The amendments identify internal services as within expanded scope.

Which date is the transition deadline for existing systems under the 2025 update?

The directive gives pre-existing systems until 24 June 2026 for new or updated requirements.