Governance and data

AI Compliance: Quebec, Ontario and the EU

Quebec privacy rules, Ontario hiring disclosure and the EU AI Act: check their scope and assign owners, controls and evidence.

Canada's AI compliance patchwork across jurisdictions

Quick answer

Check the activity and legal trigger for each AI use. Quebec automated-decision and PIA rules, Ontario job-posting disclosures and EU requirements have distinct conditions. Assign an owner, the required control and evidence for each applicable obligation.

Want to put these questions into practice with your team?

Explore AI workshops

AI governance in Canada requires more than following one federal policy announcement. Bill C-27, which included AIDA, was not adopted before the parliamentary session ended on January 6, 2025. Canada has since launched its AI for All strategy, on June 4, 2026. A strategy is not itself a new AI statute. Parliament’s record and the official strategy announcement establish those two different milestones.

The practical task is to identify which existing rules apply to your activities. Quebec privacy obligations, Ontario hiring disclosures and the EU AI Act have different triggers. An inventory helps you ask the right questions; the fact that a tool uses AI does not settle its legal scope.

Quebec: two obligations to distinguish

Section 12.1 of Quebec’s private-sector privacy law concerns a business using personal information to make a decision based exclusively on automated processing of that information. It requires notification no later than communication of the decision. It also provides for information on request, including the reasons, principal factors and parameters, and the right to have the information used corrected. The person must be able to submit observations to a staff member in a position to review the decision. Read section 12.1.

A hiring tool, credit model or account-management assistant is a starting point for investigation, not an automatic legal classification. Document what it does, how personal information is used and who actually decides.

Section 3.3 concerns a different question: projects to acquire, develop or overhaul an information system or electronic service involving personal information. The privacy officer must be consulted from the outset and the PIA must be proportionate. The CAI also explains other situations requiring a PIA. Its guidance distinguishes those triggers, in French. An elevated-risk label alone is not the wording of section 3.3.

My operational concern is the handoff between IT, legal and the people running the process. Who identifies a new use? Who examines the applicable requirements? Who implements the controls? Those responsibilities need an owner even when the software comes from a supplier.

Ontario: disclosure in covered job postings

Since January 1, 2026, covered publicly advertised Ontario job postings must include a statement disclosing AI used to screen, assess or select applicants. Employers with fewer than 25 employees on the posting day are exempt; the definition of a covered posting and other exceptions also matter. Ontario’s guide and Regulation 476/24 set out the conditions.

For a team recruiting across provinces, I would keep the posting, the recruiting workflow and the applicable disclosure together. Assign someone to check the rule before publication, including when a third party posts on your behalf. A common internal template can help, but it still needs the correct scope and facts for each use.

EU AI Act: check the requirements that apply to your use

The EU AI Act became generally applicable on August 2, 2026, with exceptions and later dates for certain obligations. The Commission’s current FAQ distinguishes those stages. It also explains that scope depends on activities such as placing systems on the EU market or putting them into service or using them in the EU. Check your role, the system’s use and its risk category against the Commission’s current guidance.

A Canadian business connection to Europe does not by itself identify every applicable obligation. If a client asks for governance evidence, clarify the system and contractual requirements involved before promising a particular compliance status.

What this means in practice

A company hiring in Ontario, handling personal information in Quebec and supplying an AI system for EU use may need several analyses. Treat them as related questions with distinct conditions, rather than assuming the strictest-sounding rule replaces the rest.

ISO/IEC 42001 can provide a management framework for responsibilities, risks, monitoring and improvement. It does not replace the legal analysis or guarantee that one exercise satisfies every jurisdiction.

I would start with one record for each use:

  • The task, the information involved and the decisions supported or made.
  • The activities and jurisdictions to examine, with a source for each applicable requirement.
  • The owner who validates the interpretation and the person who implements the control.
  • The evidence to retain: notice, assessment, review record or completed check, as appropriate.
  • The next review point, including a change in the tool, data, workflow or law.

This is a proposed working method, not a universal statutory checklist. It makes unresolved questions visible and lets you track corrections. For a closer look at everyday privacy questions, see Loi 25 and AI at work. For the operating discipline behind the records, see what quality auditing taught me about AI governance.

A documented system can support future work. Any new requirements still need to be examined when adopted; preparation cannot guarantee compliance with a future law.

Sources

Sources and scope. Official references are linked beside the relevant passages. Adapt the proposed working methods to your tasks, information and responsibilities.

Start with your team’s work

Explore custom workshops built around relevant situations and appropriate human review.

Explore AI workshops

Back to the articles