Skip to content

Guide · Technology, Privacy and AI

Artificial Intelligence Law in Canada

Artificial intelligence is regulated in Canada without a dedicated AI statute. This guide explains what happened to AIDA, how privacy law and Quebec's automated-decision rules apply, Ontario's AI disclosure in job postings, OSFI's model risk guideline E-23, liability for AI outputs, unresolved IP questions and how to build a governance programme now.

10 min read Updated
ENFR

Key takeaways

  • The proposed Artificial Intelligence and Data Act died with Bill C-27 on prorogation in January 2025.
  • Quebec requires notice when decisions are based exclusively on automated processing, with reasons on request.
  • Ontario employers with 25+ employees must disclose AI use in job postings from 1 January 2026.
  • OSFI Guideline E-23 on model risk, covering AI and machine learning, takes effect on 1 May 2027.
  • Businesses remain liable for AI outputs under negligence, misrepresentation and consumer protection law.

Artificial intelligence law in Canada: the current position

Canada does not have a statute dedicated to regulating artificial intelligence. The proposed Artificial Intelligence and Data Act (AIDA), part of Bill C-27, reached committee in the 44th Parliament but died when Parliament was prorogued on 6 January 2025. The federal government now has a Minister of Artificial Intelligence and Digital Innovation, who sponsors the current privacy reform bill, but no standalone federal AI bill had been introduced in the current Parliament as of the sources reviewed for this page in November 2026.

That does not leave artificial intelligence unregulated. Privacy, human rights, employment, consumer protection, competition, financial services, product liability and intellectual property laws all apply to how AI systems are built, bought and used.

Neural network visualization on a screen representing artificial intelligence regulation and governance in Canada
Without an AI-specific statute, Canadian businesses govern AI through privacy, employment, consumer and sector rules.

For businesses, the practical question is therefore not "what does the AI law require?" but "which existing obligations does this AI use case trigger, and how do we show we have met them?". This page maps the main sources of obligation, from Quebec's rules on automated decisions and Ontario's hiring disclosure requirement to OSFI's model risk guideline, explains liability and intellectual property issues, and sets out the components of an AI governance programme that will also prepare the business for future legislation.

Privacy law and automated decisions

Privacy law is the most direct constraint on artificial intelligence in Canada, because most commercially significant AI systems use personal information to train, fine-tune, operate or evaluate models. Under PIPEDA and the provincial private-sector statutes, organizations need a valid basis, usually meaningful consent, for collecting and using personal information for AI purposes; must limit collection to what is necessary; must be transparent about purposes; and must protect information used in training and inference. Using customer data collected for one purpose to train a model for a different purpose is a new use that may require fresh consent.

Quebec has the most specific rules. Under s. 12.1 of its private-sector privacy act, an enterprise that uses personal information to render a decision based exclusively on automated processing must inform the person concerned no later than when it informs them of the decision. On request, it must also tell the person what personal information was used, the reasons and principal factors and parameters that led to the decision, and of their right to have the information corrected.

The person must be given an opportunity to submit observations to a member of the enterprise's personnel who can review the decision. Separately, s. 8.1 requires advance notice when technology that can identify, locate or profile a person is used.

The successor federal privacy bill, Bill C-36, is at second reading in the House of Commons according to LEGISinfo. Businesses should expect any new federal privacy law to address automated decision systems and transparency, building on concepts already in Quebec law. Our insight on automated decisions under Law 25 explains the Quebec rules in practice.

Employment, human rights and AI in hiring

AI tools for recruitment, scheduling, performance management and monitoring raise employment and human rights issues. Since 1 January 2026, Ontario employers with 25 or more employees that publicly advertise job postings must disclose in the posting whether artificial intelligence is used to screen, assess or select applicants, according to the Ontario government's guide to the Employment Standards Act, 2000. The same rules require postings to include expected pay or a pay range, state whether a vacancy exists and not require Canadian experience.

Beyond disclosure, human rights statutes in every jurisdiction prohibit discrimination in employment on protected grounds. An AI screening tool that disproportionately disadvantages applicants on a protected ground can expose the employer to a discrimination complaint even if the bias was unintentional and the tool was supplied by a vendor. Employers should test tools for adverse impact, keep humans meaningfully involved in decisions and require vendors to support audits. Our insight on AI in hiring in Canada covers these obligations in detail.

AI is also used after hiring, for scheduling, productivity analytics, performance evaluation and workplace monitoring. These uses engage privacy law, collective agreements in unionized workplaces and, where decisions about discipline or termination rely on AI outputs, the employer's obligation to act fairly and on accurate information. Employers should be able to explain how an AI-assisted decision was reached, give employees a way to challenge it and ensure that managers do not treat outputs as conclusive. In Quebec, decisions based exclusively on automated processing trigger the notice and explanation rules described above.

Financial services and other sector rules

Regulated sectors face additional expectations. OSFI's Guideline E-23 on model risk management, which takes effect on 1 May 2027, applies to models used by federally regulated financial institutions, including artificial intelligence and machine learning models. It expects institutions to maintain an enterprise-wide model risk management framework covering model inventory, risk rating, development, validation, approval, monitoring and decommissioning, with governance proportionate to the model's risk. In Quebec, the AMF has also addressed the use of AI by financial institutions; see our insight on the AMF AI guideline.

Other regulators apply existing powers. Health Canada regulates software as a medical device, including AI-enabled diagnostic and treatment tools; securities regulators expect registrants and issuers to manage and disclose AI-related risks accurately; and the Competition Bureau has signalled interest in how algorithms affect pricing and competition. Our insight on software as a medical device covers the health sector.

Source of obligationWhat it requiresStatus (November 2026)
PIPEDA and provincial privacy lawsConsent, purpose limitation, transparency and safeguards for personal information in AI systemsIn force
Quebec private-sector act, s. 12.1Notice and explanation of decisions based exclusively on automated processingIn force
Ontario ESA job posting rulesDisclose AI use in screening, assessing or selecting applicants (25+ employees)In force since 1 January 2026
OSFI Guideline E-23Model risk management for FRFIs, including AI/ML modelsEffective 1 May 2027
Voluntary Code of Conduct on advanced generative AIAccountability, safety, fairness, transparency, human oversight, robustnessVoluntary (September 2023)
AIDA (Bill C-27)Proposed risk-based AI regulationDied on prorogation, January 2025
Practical point

Start with an inventory. Many organizations do not know where AI is already used, often embedded in vendor software. A register of AI use cases, the data each uses, who owns it and which decisions it affects is the foundation for privacy assessments, Quebec automated-decision notices, Ontario hiring disclosures, OSFI model inventories and any future statutory obligations.

Liability, consumer protection and AI outputs

Businesses remain responsible for what their AI systems say and do. In Moffatt v. Air Canada, a 2024 decision of British Columbia's Civil Resolution Tribunal, an airline was held liable for negligent misrepresentation after its website chatbot gave a customer incorrect information about bereavement fares; the tribunal rejected the argument that the chatbot was a separate entity responsible for its own statements. The decision is from a small-claims tribunal and is not binding precedent, but it illustrates how ordinary negligence and misrepresentation principles apply. See our insight on AI chatbots and liability.

Other liability routes include:

  • Misleading advertising: claims about what an AI product can do, or that a product uses AI when it does not, can engage the deceptive marketing provisions of the Competition Act and provincial consumer protection statutes.
  • Product liability: where AI is embedded in physical products, such as vehicles, devices or machinery, defects can lead to negligence claims, and in Quebec to the Civil Code's safety-defect regime.
  • Contract: service levels, warranties and limitation clauses in AI contracts determine who bears the cost of errors between businesses.
  • Defamation and privacy torts: generated content about real people can give rise to claims.

Intellectual property and AI

Canada's Copyright Act has not been amended to address AI training or AI-generated works. The government has consulted on copyright in the age of generative AI, but questions remain open, including whether training on copyrighted works requires authorization or falls within an exception such as fair dealing, and whether works generated with little human input can be protected at all, given that Canadian copyright has traditionally required an exercise of skill and judgment by a human author. Patent law raises parallel questions about AI-assisted inventions and the patentability of AI-related software. Our insights on AI training data copyright risk and AI-generated works discuss the issues, and our intellectual property page provides an overview.

In the meantime, contracts carry the risk allocation. Agreements with AI vendors should address ownership of inputs and outputs, restrictions on using customer data to train vendor models, IP infringement indemnities and how they apply to generated content, confidentiality, security and audit rights. See our insight on AI vendor contracts.

Generative AI in the workplace

Employee use of generative AI tools is often the first AI risk a business has to manage, because staff adopt public tools before any formal programme exists. The main legal risks are disclosure of confidential or personal information through prompts, reliance on inaccurate outputs, infringement of third-party rights in generated content and failure to disclose AI use where clients, courts or regulators expect it. Canadian courts, for example, have issued practice directions on the use of AI in court submissions, and professionals in regulated fields face their own conduct rules.

A workable employee policy usually:

  • identifies approved tools and prohibits entering confidential, personal or client information into unapproved ones;
  • requires human review of outputs before they are relied on or sent externally;
  • sets rules for disclosure of AI assistance to customers and in published work;
  • addresses intellectual property ownership and checks for infringing content; and
  • explains how employee use may be monitored, consistent with privacy law and, in Ontario, the employer's electronic monitoring policy.

Our insight on generative AI policies for employees provides a fuller checklist.

AI in transactions and preparing for legislation

Artificial intelligence is now a standard diligence topic in acquisitions and investments, particularly in technology businesses whose value depends on models and data. Buyers should ask how training data was sourced and whether its use was authorized, whether personal information was used lawfully, which open-source models and licences are involved, who owns model weights and outputs, what testing and incident records exist, and whether customer contracts restrict the use of their data. Gaps can be addressed through specific representations, indemnities, remediation covenants or price.

Future legislation is likely, even if its timing and form are uncertain. Draft frameworks in Canada and abroad, including the former AIDA, the European Union's AI Act and international standards, share common features: risk classification, impact assessments, documentation, human oversight, incident reporting and transparency to affected individuals. Businesses that build these elements into their governance now will be able to adapt to a new Canadian statute with limited disruption, and will already meet the expectations of many customers, investors and foreign regulators.

Building an artificial intelligence governance programme

An artificial intelligence governance programme lets a business use AI with confidence now and adapt quickly when legislation arrives. The federal government's Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems, published in September 2023, offers a useful structure, built around accountability, safety, fairness and equity, transparency, human oversight and monitoring, and validity and robustness. International standards such as ISO/IEC 42001 provide a certifiable management-system approach. Core elements include:

  1. Policy and accountability: an AI policy approved by senior management, with named owners for each system and board reporting on material AI risks.
  2. Inventory and risk classification: a register of AI use cases rated by impact on individuals, customers and the business.
  3. Impact assessments: privacy, human rights and fairness assessments for higher-risk uses, integrated with privacy impact assessments.
  4. Data governance: lawful sourcing of training data, documentation of datasets and controls on personal and confidential information.
  5. Testing and monitoring: pre-deployment testing for accuracy, bias and security, and monitoring for drift and incidents after deployment.
  6. Transparency: notices to users and affected individuals, including Quebec automated-decision notices and Ontario job posting disclosures.
  7. Human oversight: clear points at which people review, override or explain AI outputs.
  8. Vendor management: diligence and contract terms for third-party AI tools.
  9. Employee use: rules for staff use of generative AI tools, covering confidential information, accuracy checks and disclosure.

Illustration: a hypothetical Canadian insurer deploys a machine learning model to triage claims. It must assess privacy consent for using claims data, give Quebec claimants notice of any decision made exclusively by automated processing and the factors on request, test the model for unfair outcomes, bring it within its model risk framework ahead of OSFI E-23's effective date, and ensure that its vendor contract restricts reuse of claims data. Our insight on board oversight of AI risk explains the governance role of directors.

How TRW & Co helps

We advise developers, deployers and investors on artificial intelligence law in Canada: mapping obligations for specific use cases, building AI governance programmes and inventories, privacy and automated-decision assessments, Ontario hiring disclosures, OSFI E-23 readiness, AI vendor and customer contracts, IP risk in training data and outputs, and AI issues in transactions. We track federal and provincial reform and help clients prepare for it. Where proceedings require Canadian-licensed counsel, we work with them closely. This page is general information, not legal advice. To discuss an AI project, please contact us.

Frequently asked questions

Does Canada have an AI law?

Not a dedicated one. The Artificial Intelligence and Data Act, proposed in Bill C-27, died when Parliament was prorogued in January 2025, and no standalone federal AI bill had been introduced in the current Parliament as of November 2026. AI is regulated through privacy, human rights, employment, consumer protection, competition, sector-specific and IP laws.

What does Quebec require for automated decisions?

Under s. 12.1 of Quebec's private-sector privacy act, an enterprise that makes a decision based exclusively on automated processing of personal information must inform the person no later than when it communicates the decision. On request, it must disclose the information used, the reasons and principal factors and parameters, and the right to correction, and allow the person to submit observations for review.

Do Ontario employers have to disclose AI use in hiring?

Yes. Since 1 January 2026, Ontario employers with 25 or more employees must state in publicly advertised job postings whether artificial intelligence is used to screen, assess or select applicants. Human rights law also prohibits discriminatory outcomes from AI tools, so employers should test tools for adverse impact.

Can a company be liable for what its chatbot says?

Yes. In Moffatt v. Air Canada, a 2024 decision of British Columbia's Civil Resolution Tribunal, an airline was held liable for negligent misrepresentation based on incorrect information given by its website chatbot. The decision is not binding precedent, but it shows that ordinary liability principles apply to AI-generated statements.

When does OSFI's E-23 guideline apply?

OSFI's Guideline E-23 on model risk management takes effect on 1 May 2027. It applies to federally regulated financial institutions and covers models including artificial intelligence and machine learning models, requiring an enterprise-wide framework for inventory, risk rating, validation, approval, monitoring and decommissioning proportionate to model risk.

Sources

  1. LEGISinfo: Bill C-27 (44th Parliament, 1st Session)
  2. Act respecting the protection of personal information in the private sector, CQLR c. P-39.1
  3. Ontario: Your guide to the ESA, requirements related to publicly advertised job postings
  4. OSFI Guideline E-23: Model Risk Management (2027)
  5. ISED: Voluntary Code of Conduct on Advanced Generative AI Systems
  6. LEGISinfo: Bill C-36 (45th Parliament, 1st Session)

General information only, not legal advice. References to statutes, courts and authorities describe published law and practice at the date shown.