Canada AIDA AI Compliance Checklist: Ten Items for Canadian AI Without a Federal AI Act
A working checklist for Canadian organizations running AI over personal information now that the Artificial Intelligence and Data Act has died with Bill C-27 and PIPEDA, Quebec Law 25, OSFI guidance and the Treasury Board Directive carry the load. Each item names the obligation, the concrete action, and the evidence it produces, so the list works as an audit-readiness pass. Ordered the way a Canadian privacy review moves: identity and accountability first, then consent and classification, then transfers, then automated decisions.
The Artificial Intelligence and Data Act never came into force. It rode inside Bill C-27, and the bill died on the Order Paper when Parliament was prorogued on 6 January 2025, with no reintroduction as of August 2026. The obligations that survived are PIPEDA, Quebec's Law 25, OSFI Guideline E-23 for federally regulated financial institutions, and the Treasury Board Directive on Automated Decision-Making for federal departments.
This checklist follows the order a Canadian privacy review tends to take: identity and accountability records first, then consent and classification, then cross-border communication, then automated decisions and breach readiness. Each item pairs the action with the evidence it produces, because PIPEDA's accountability principle asks an organization to demonstrate rather than assert.
1. Bind identity to every AI request
Route AI traffic so a named user or agent, and not a shared service key, is attached to each prompt. Evidence is a per-request record showing which identity sent which prompt to which model, which is what PIPEDA Principle 4.1 accountability rests on when a reviewer asks who did this.
2. Classify the personal information in the prompt
Evaluate the classification of the context window before it leaves, separating ordinary personal information from the sensitive categories that raise the consent and safeguard bar under PIPEDA Principles 4.3 and 4.7. Evidence is a per-request classification decision recorded at the boundary.
3. Enforce the consent basis by role
Block prompts carrying sensitive personal information when the caller's role has no basis for that use or disclosure. Evidence is the per-request policy decision showing exactly what each role was permitted to send into a model.
4. Record the destination region of every transfer
Log the model endpoint and its region for each request carrying personal information. Evidence is a destination log that answers the PIPEDA comparable-protection question and the Law 25 outside-Quebec communication question directly, rather than through a vendor list.
5. Enforce cross-border rules at the boundary
Route prompts carrying regulated classifications only to endpoints in permitted regions or under an approved arrangement. Evidence is the enforcement log, and AI data residency controls are how this becomes operational rather than contractual.
6. Produce a per-decision record for automated decisions
For any AI output that decides something about an individual, write a record binding the input prompt to the output and the applied policy. Evidence is decision lineage, the operational backing for Quebec Law 25 section 12.1, which has been in force since 22 September 2023 and mirrors GDPR Article 22 handling.
7. Support the section 12.1 review right
Flag decisions based exclusively on automated processing so you can inform the individual, supply the personal information used and the principal factors behind the decision, and route the case to a person for review. Evidence is the review-flag record tied to the specific decision being contested.
8. Replace static API keys with per-role, per-route policy
Remove shared credentials that grant permanent full model access to any caller, which is where PIPEDA Principle 4.7 safeguards and OSFI Guideline E-23 model risk expectations meet. Evidence is an access policy plus the enforcement log showing which roles reached which models.
9. Keep the record independent of the application
Ensure the compliance log is written by a system separate from the one making the AI call, so it survives selective logging, suppression, and crash-loss. Evidence is a signed, tamper-evident log outside application custody, as argued in You Own the AI Liability, Not the Vendor.
10. Pre-build the breach-reporting path
Confirm you can reconstruct an AI-channel exposure, because PIPEDA requires reporting a breach of security safeguards to the Privacy Commissioner and notifying individuals where there is a real risk of significant harm. Evidence is a query against the independent request log that produces a timeline. My honest take: teams that skipped this discover during the incident that the risk test is unanswerable without prompt-level records.
How the checklist maps to the controls
Items 1 through 10 resolve to a handful of enforcement points at the request boundary, laid out in the Canada AIDA controls mapping. The audit-evidence view covers what each artifact looks like when the Office of the Privacy Commissioner or the Commission d'accès à l'information asks. Together they turn this list into a running system with evidence attached.
DeepInspect
Nine of these ten items produce evidence at one place: the boundary between your users or agents and the LLM APIs they call. DeepInspect sits inline at that boundary. It binds identity to each request, classifies the prompt, enforces consent-basis and destination policy, flags automated decisions for review, and writes a signed per-decision record before the response returns. The checklist becomes a query rather than a scramble.
For a Canadian organization that shelved its AI governance work when AIDA collapsed, items 1 and 7 are the two that a Quebec resident's request or a PIPEDA complaint will hit first. Book a technical deep dive at deepinspect.ai.
Frequently asked questions
- Is this checklist enough for Canadian AI compliance?
It covers the AI-traffic obligations a Canadian privacy review reaches for: accountability, consent basis, safeguards, cross-border communication, automated decisions, and breach reconstruction. It does not replace a full privacy programme covering collection notices, retention schedules, the privacy officer role, and access requests across all channels. Treat it as the AI-specific layer on top of an existing programme.
- What changed when AIDA died?
The proposed risk-tiering, the AI and Data Commissioner, and the statutory record-keeping duty for high-impact systems all went away with Bill C-27 in January 2025. Nothing was removed from PIPEDA or Law 25, so the accountability, safeguard, and automated-decision obligations that predate AIDA continue to apply exactly as they did.
- Does Quebec Law 25 apply to organizations outside Quebec?
Law 25 applies to the personal information of Quebec residents handled by organizations carrying on an enterprise in Quebec, so a company headquartered elsewhere can be caught by section 12.1 through its Quebec customers or employees. Scoping this by where the individual lives rather than where the servers sit is the safer reading.
- Will a future Canadian AI act invalidate this work?
The four artifacts here, identity binding, prompt classification, destination logging, and per-decision records, are the base layer any risk-tiered regime builds on. A successor statute would add classification duties and reporting on top rather than replace the underlying records, so the checklist grows rather than resets.