AI regulation in Canada — privacy law, provincial rules and the status of federal AI legislation

Appropriate purposes s. 5(3)

An organization may collect, use or disclose personal information only for purposes that a reasonable person would consider appropriate in the circumstances.

Evidence commonly relevant: Purpose statement for each AI use of personal data; Reasonable-person assessment

Accountability Schedule 1, Principle 4.1

An organization is responsible for personal information under its control and must designate an individual accountable for compliance, including for information transferred to third parties for processing.

Evidence commonly relevant: Named privacy officer; Vendor/processor contracts for AI services; Privacy management program

Knowledge and consent Schedule 1, Principle 4.3; s. 6.1

Knowledge and consent of the individual are required for collection, use or disclosure of personal information except where inappropriate; consent is valid only if the individual can reasonably understand the nature, purpose and consequences of the processing.

Evidence commonly relevant: Consent language covering AI uses (training, profiling); Consent records

Safeguards Schedule 1, Principle 4.7

Personal information must be protected by security safeguards appropriate to its sensitivity (physical, organizational and technological), including against loss, theft, unauthorized access, disclosure, copying, use or modification.

Evidence commonly relevant: Security controls for AI data pipelines and prompts; Access control records

Openness Schedule 1, Principle 4.8

Organizations must make information about their personal-information policies and practices readily available in an understandable form, including a contact for inquiries.

Evidence commonly relevant: Public privacy policy describing AI uses

Individual access and correction Schedule 1, Principle 4.9

On request, individuals must be told whether the organization holds their personal information, be given access to it and an account of its use and disclosure, and be able to challenge accuracy and have it amended.

Evidence commonly relevant: Access-request procedure covering AI-derived data; Correction logs

Legal authority and meaningful consent for generative AI uses Principle 1from 2023-12-07

Organizations must have a lawful basis (including valid, meaningful consent where relied upon) for collecting and using personal information to develop, train, deploy or use generative AI, including for web-scraped training data.

Evidence commonly relevant: Lawful-basis assessment for training and prompting data; Consent records

Openness, accountability and explainability for generative AI Principles 4-5from 2023-12-07

Organizations should be transparent about how personal information is handled by generative AI, disclose when AI is involved in interactions or decisions, and be accountable through governance, impact assessments, complaint handling and the ability to explain outputs that affect individuals.

Evidence commonly relevant: Privacy impact assessment for the generative AI use; AI disclosure notices; Governance records

AIDA proposed high-impact system duties (not in force) Part 3 of Bill C-27 (proposed AIDA)

The proposed AIDA would have required persons responsible for high-impact AI systems to identify and mitigate risks of harm and bias, keep records, publish plain-language descriptions, monitor systems and report material harm. None of this is law; it is listed only so products do not present it as a current obligation.

Evidence commonly relevant: None required

Proposed: transparency about automated decision making for significant decisions Bill C-36 (clause numbers not verified)

If enacted, organizations would have to be transparent about automated decision making used for significant decisions about individuals, alongside stronger consent, children's-data, deletion and cross-border-transfer rules. Not in force.

Evidence commonly relevant: Inventory of automated significant decisions (readiness)

Transparency: publish system information and identify AI content Transparency principlefrom 2023-09-01

Signatories commit to publishing information on capabilities and limitations, ensuring AI-generated content can be identified (e.g. watermarking), and disclosing when people interact with the system.

Evidence commonly relevant: Published capability/limitation statement; Content-labelling mechanism

Safety, human oversight, validity and robustness measures Safety; Human oversight and monitoring; Validity and robustness principlesfrom 2023-09-01

Signatories commit to risk assessment and mitigation before deployment, adversarial testing (red-teaming), monitoring after release with incident reporting, and benchmarking for validity and robustness.

Evidence commonly relevant: Risk assessment; Red-team reports; Post-deployment monitoring records

Person in charge of the protection of personal information s. 3.1

The person exercising the highest authority in the enterprise is by default in charge of personal information protection and may delegate in writing; the title and contact information must be published on the enterprise's website.

Evidence commonly relevant: Written delegation; Website publication of title and contact

Governance policies and practices s. 3.2

Enterprises must establish and implement governance policies for personal information (retention and destruction, roles and responsibilities, complaint process), proportionate to their activities, approved by the privacy officer and published in clear language.

Evidence commonly relevant: Published governance policy; Privacy officer approval

Privacy impact assessment for information-system projects s. 3.3 (and s. 17 for transfers outside Quebec)

A PIA is mandatory for any project to acquire, develop or overhaul an information system or electronic service delivery system involving personal information (which captures most AI deployments), consulting the privacy officer from the outset and proportionate to sensitivity, purpose, quantity and medium; a PIA is also required before communicating personal information outside Quebec.

Evidence commonly relevant: Completed PIA per AI project; Cross-border transfer assessment and written agreement

Prior notice for technology that identifies, locates or profiles s. 8.1

When collecting personal information using technology with functions allowing a person to be identified, located or profiled (profiling includes assessing work performance, economic situation, health, preferences, interests or behaviour), the enterprise must first inform the person of the technology and of how to activate those functions.

Evidence commonly relevant: Notice text; Default-off configuration evidence

Decision based exclusively on automated processing s. 12.1

An enterprise using personal information to render a decision based exclusively on automated processing must inform the person no later than when informing them of the decision; on request it must disclose the personal information used, the reasons and principal factors and parameters that led to the decision, and the right to have that information corrected; and it must give the person an opportunity to submit observations to a staff member able to review the decision. Failure to inform is subject to monetary administrative penalties (s. 90.1(5)).

Evidence commonly relevant: Decision notices; Explanation template (factors/parameters); Human-review request log

Responsibility, designated individual and consent ss. 5, 7, 8

Organizations are responsible for personal information in their custody or control (including when handled by service providers), must designate one or more individuals responsible for compliance, and may only collect, use or disclose personal information with consent unless the Act provides otherwise.

Evidence commonly relevant: Designated privacy contact; Consent records for AI uses; Service-provider contracts

Reasonable safeguards and breach notification ss. 34, 34.1

Organizations must make reasonable security arrangements against unauthorized access, collection, use, disclosure, copying, modification, disposal or destruction, and must notify the Commissioner without unreasonable delay of a loss or unauthorized access/disclosure where a real risk of significant harm exists.

Evidence commonly relevant: Security controls over AI data; Breach-response procedure

Designated individual, policies and complaint process ss. 4, 5

Organizations must designate one or more individuals responsible for compliance, make their position titles and contact information publicly available, develop and follow policies and practices necessary to meet the Act, and have a complaint-response process.

Evidence commonly relevant: Published privacy contact; Policies covering AI use of personal data

Valid consent and reasonable security arrangements ss. 6-8, 34

Collection, use and disclosure require consent (express, implied or opt-out with notice, as permitted); consent cannot be required beyond what is necessary to provide the product or service and is invalid if obtained by deception. Organizations must make reasonable security arrangements against unauthorized access, collection, use, disclosure, copying, modification or disposal.

Evidence commonly relevant: Consent notices referencing AI uses; Security controls

Disclose AI use in publicly advertised job postings ESA 2000 s. 8.4(1); O. Reg. 476/24 ss. 1–2from 2026-01-01

An employer that advertises a publicly advertised job posting and uses artificial intelligence to screen, assess or select applicants must include in the posting a statement disclosing that AI is used. Employers with fewer than 25 employees on the day the posting is posted are exempt, and postings for work performed outside Ontario are excluded.

Evidence commonly relevant: Copy of job postings showing the AI-use statement; Inventory entry for the screening tool; Record of which postings used AI screening

Definition of artificial intelligence and posting exemptions O. Reg. 476/24 ss. 1, 2(1)from 2026-01-01

"Artificial intelligence" means a machine-based system that, for explicit or implicit objectives, infers from the input it receives to generate outputs such as predictions, content, recommendations or decisions that can influence physical or virtual environments. General recruitment campaigns, help-wanted signs, internal-only postings and postings for work performed outside Ontario are not publicly advertised job postings. Part III.1 does not apply to employers with fewer than 25 employees on the posting day.

Evidence commonly relevant: Headcount on posting date; Posting classification record