1. Introduction

AI accountability in Canadian civil law concerns the legal responsibility of individuals, businesses, developers, employers, public authorities, and other organisations for harm caused by artificial intelligence (AI) systems.

AI systems are increasingly used in healthcare, financial services, employment recruitment, insurance, education, transportation, policing, content creation, and consumer products. Although these systems can improve efficiency, they can also produce inaccurate information, discriminate against individuals, infringe privacy, generate defamatory statements, cause financial losses, or contribute to physical injury.

Canadian civil law provides several possible avenues for addressing such harm, including negligence, product liability, defamation, breach of contract, privacy-related claims, and claims based on statutory human rights protections.

For example, if a Canadian employer uses an AI recruitment tool that systematically excludes qualified applicants because of a protected characteristic, several legal questions arise:

Who selected and deployed the system?

Was the tool reasonably tested for discriminatory outcomes?

Did the employer rely on its results without meaningful human review?

Can the affected applicant prove a legally recognised injury or loss?

Does the conduct violate applicable provincial human rights legislation?

Is the developer, vendor, employer, or another party legally responsible?

AI accountability does not mean that every harmful AI output automatically creates a legal claim. A claimant must establish the elements of a recognised cause of action, satisfy applicable procedural requirements, and identify a legally responsible defendant.

Important distinction: Canada does not have a single comprehensive civil-liability statute governing every AI-related injury. Liability is assessed through existing common-law principles, Quebec's civil-law framework, legislation, and the particular facts of each case.

The cases discussed below include decisions directly concerning AI-generated material and decisions establishing broader Canadian principles that may apply to AI disputes. Where a decision is not specifically an AI case, that limitation is identified.

2. Meaning and scope of AI accountability

AI accountability refers to the legal and practical responsibility for developing, supplying, deploying, supervising, and using AI systems.

Civil claims may arise where an AI system:

Produces false or misleading information about an identifiable person.

Discriminates in recruitment, lending, housing, insurance, or public services.

Discloses personal or confidential information without lawful authority.

Causes financial loss through negligent design or implementation.

Contributes to physical injury through unsafe automated operation.

Breaches contractual promises about accuracy, security, or performance.

Uses protected content in circumstances giving rise to an actionable intellectual-property claim.

Makes consequential decisions without safeguards required by law.

The central legal question is not simply whether AI was involved. It is whether the conduct of a particular person or organisation breached a legal obligation and caused a legally compensable harm.

Principal categories of civil AI claims

A. Negligence claims

Claims alleging that a developer, supplier, employer, or user failed to take reasonable care in designing, testing, selecting, deploying, or supervising an AI system.

B. Discrimination and human rights claims

Claims involving biased automated recruitment, credit decisions, facial recognition, eligibility assessments, or other AI-assisted decisions that adversely affect protected groups.

C. Privacy and confidentiality claims

Claims involving unlawful collection, use, disclosure, retention, or exposure of personal information by AI systems.

D. Defamation and misinformation claims

Claims arising when an AI system generates or distributes a false statement that harms a person's reputation and satisfies the legal requirements for defamation.

3. Canadian legal framework for AI civil liability

A. Common-law provinces

In provinces such as Ontario, British Columbia, and Alberta, civil AI claims generally rely on established common-law causes of action, subject to provincial legislation.

Relevant doctrines include:

Negligence and the duty of care.

Contractual liability.

Defamation.

Product liability and negligent misrepresentation.

Breach of confidence.

Privacy-related torts recognised in the relevant province.

Vicarious liability for conduct sufficiently connected to an employment relationship.

A claimant must establish the elements of the specific cause of action. A general allegation that an AI system was unfair, inaccurate, or harmful is not sufficient by itself.

B. Quebec's civil-law framework

Quebec uses a civil-law system based principally on the Civil Code of Québec.

Articles 1457 and 1458 are particularly relevant:

Article 1457 establishes general civil liability for a person who fails to comply with applicable standards of conduct and causes injury to another.

Article 1458 governs contractual civil liability where a person fails to perform contractual obligations.

Depending on the facts, Quebec AI litigation may therefore involve fault, injury, causation, contractual obligations, and statutory duties under the Civil Code and other applicable legislation.

The common-law tort of negligence should not simply be imported into Quebec without accounting for its distinct legal framework.

C. Canadian Charter of Rights and Freedoms

The Charter may be relevant when AI is used by governments or other actors whose conduct is subject to the Charter.

Potential issues include:

Equality rights under section 15.

Life, liberty, and security of the person under section 7.

Freedom of expression under section 2(b).

Procedural fairness and the legality of public decision-making.

The Charter generally applies to government action rather than purely private conduct. A private company's use of AI does not automatically create a Charter claim.

D. Privacy legislation

AI systems often process large quantities of personal information. Relevant Canadian privacy laws may include:

The federal Personal Information Protection and Electronic Documents Act (PIPEDA), where applicable.

Provincial private-sector privacy legislation.

Public-sector freedom-of-information and privacy legislation.

Quebec's Act respecting the protection of personal information in the private sector.

British Columbia's and Alberta's private-sector privacy legislation.

The applicable legislation depends on the organisation, activity, jurisdiction, and type of information involved. Some statutes provide regulatory complaint procedures or specific remedies rather than an unrestricted right to sue for every privacy violation.

E. Human rights legislation

Federal and provincial human rights laws prohibit discrimination in specified areas, including employment and services, where the relevant statutory conditions are satisfied.

An AI tool that produces discriminatory outcomes may expose an employer or service provider to a human rights complaint even if the discrimination was not intentional.

The organisation's reliance on an external vendor does not necessarily eliminate its own obligations to affected individuals.

F. Consumer protection and product liability

AI products and services may also be subject to consumer protection laws, contractual obligations, and product liability principles.

Potential claims can concern:

Misleading representations about AI capabilities.

Failure to disclose important limitations.

Defective products or unsafe systems.

Breach of express or implied contractual terms.

Inadequate warnings or instructions.

The existence of a product defect, the applicable standard of care, and the available remedy must be determined under the relevant provincial or federal law.

4. Important Canadian case laws on AI accountability and related civil liability

A key point for legal research is that Canadian courts have relatively limited direct case law specifically addressing AI liability. The following decisions include one important chatbot-liability ruling and several authoritative cases concerning online platforms, privacy, defamation, and technological accountability that may inform future AI civil claims.

Case 1: Moffatt v. Air Canada (2024)

British Columbia Civil Resolution Tribunal · 2024 BCCRT 149

Facts: Jake Moffatt used Air Canada's website chatbot to obtain information about bereavement fares after the death of his grandmother. The chatbot indicated that a passenger could apply for a bereavement fare retroactively within 90 days of ticket issuance. Air Canada's published policy contained a different rule. After purchasing tickets and relying on the chatbot's information, Moffatt sought a fare adjustment, which Air Canada refused.

Legal issue: Could Air Canada be held responsible for misleading information provided by its own customer-service chatbot?

Decision: The Civil Resolution Tribunal found Air Canada liable for the misleading information and awarded Moffatt approximately CAD $812.02. The tribunal rejected the argument that the company could avoid responsibility simply because the incorrect information came from a chatbot.

Legal principles:

Businesses may be held accountable for representations made through automated customer-service systems.

A company cannot necessarily avoid liability by distinguishing between information supplied by a human representative and information supplied by its chatbot.

The presence of correct information elsewhere on a website does not automatically eliminate the consequences of a misleading representation.

Application to AI civil claims: This case is particularly important for businesses using generative AI in customer support, financial services, insurance, travel, or online retail. A company that allows customers to rely on an AI assistant may face liability if its representations satisfy the elements of an actionable misrepresentation.

Limitation: This is a Civil Resolution Tribunal decision, not a Supreme Court of Canada precedent binding every Canadian court. Its reasoning is persuasive rather than universally binding.

Case 2: Douez v. Facebook, Inc. (2017)

Supreme Court of Canada · 2017 SCC 33 · [2017] 1 SCR 751

Facts: Facebook used members' names and profile photographs in advertising through its Sponsored Stories feature. Deborah Douez brought a claim under British Columbia's Privacy Act. Facebook sought to rely on a contractual forum-selection clause requiring the dispute to be litigated in California.

Legal issue: Should the court enforce an online platform's forum-selection clause where the claim concerns privacy rights and the parties have unequal bargaining power?

Decision: The Supreme Court of Canada allowed the appeal and declined to enforce the forum-selection clause in the circumstances. The majority considered the inequality of bargaining power, the consumer context, and the importance of privacy rights under British Columbia law.

Legal principles:

Standard-form online contracts are not necessarily conclusive in every dispute.

Privacy rights can carry significant public importance.

Courts may examine bargaining inequality and the public interest when deciding whether to enforce contractual clauses that restrict where a claim may be brought.

Application to AI civil claims: A person whose image or personal information is used by an AI-powered platform may face similar questions about consent, privacy, online contractual terms, and the appropriate forum for litigation.

Limitation: Douez concerned online advertising and privacy, not generative AI. It is relevant by analogy to AI platforms and their user agreements.

Case 3: R. v. Bykovets (2024)

Supreme Court of Canada · 2024 SCC 6

Facts: Police investigating fraudulent online purchases obtained internet protocol (IP) addresses from a payment-processing company. Those addresses were subsequently used to identify individuals through their internet service provider.

Legal issue: Does an individual have a reasonable expectation of privacy in an IP address for purposes of section 8 of the Canadian Charter of Rights and Freedoms?

Decision: The Supreme Court held that an IP address attracts a reasonable expectation of privacy and that police requests to obtain such information engage section 8's protection against unreasonable search and seizure.

Legal principles:

Digital identifiers may reveal sensitive information about individuals.

Privacy analysis must account for the information that technology enables others to infer.

Information that appears innocuous in isolation may provide access to much more revealing personal data.

Application to AI civil claims: AI systems can infer identities, behaviour, preferences, and other personal characteristics from digital information. The reasoning in Bykovets may inform privacy analysis involving AI-driven profiling, although the judgment itself concerns a constitutional search by police rather than a private civil damages claim.

Limitation: The case does not establish a general right to damages against private AI companies. Charter rights and private-sector privacy obligations operate under different legal frameworks.

Case 4: Jones v. Tsige (2012)

Ontario Court of Appeal · 2012 ONCA 32 · 108 OR (3d) 241

Facts: Sandra Jones discovered that a bank employee, Tsige, had repeatedly accessed her personal banking information without authorisation. The parties were employed by the same bank, and Tsige had accessed the records for personal reasons.

Legal issue: Did Ontario common law recognise a civil cause of action for the intentional intrusion into another person's private affairs?

Decision: The Ontario Court of Appeal recognised the tort of intrusion upon seclusion. It held that a person may be liable for an intentional or reckless intrusion into another person's private affairs without lawful justification, where the intrusion would be highly offensive to a reasonable person and causes the type of injury recognised by the tort.

Legal principles:

Unauthorised access to private information may give rise to civil liability.

A claimant need not always prove a financial loss to obtain damages for intrusion upon seclusion.

The nature, extent, and offensiveness of the intrusion matter.

Application to AI civil claims: An organisation that deploys an AI tool to access confidential employee files, banking records, medical information, or private communications without lawful justification may face privacy-related claims depending on the facts and applicable law.

Limitation: The tort recognised in Jones is an Ontario common-law development. It should not be assumed to apply identically in every province or to every form of automated data processing.

Case 5: Grant v. Torstar Corp. (2009)

Supreme Court of Canada · 2009 SCC 61 · [2009] 3 SCR 640

Facts: The case arose from a newspaper article concerning a prominent businessperson and allegations about a proposed development project. The dispute required the Supreme Court to consider the balance between protecting reputation and enabling responsible public-interest journalism.

Legal issue: Should Canadian defamation law recognise a defence for responsible communication on matters of public interest?

Decision: The Supreme Court recognised the defence of responsible communication on matters of public interest. The defence examines whether publication concerned a public-interest matter and whether the publisher acted responsibly in attempting to verify the information.

Legal principles:

Reputation is legally protected.

Public-interest communication must be assessed in context.

The steps taken to verify information may be relevant to a defence.

Defamation law balances protection against false statements with freedom of expression.

Application to AI civil claims: If a business publishes AI-generated allegations about an identifiable individual, the resulting dispute may involve ordinary defamation principles. Relevant questions include whether the statement was defamatory, whether it was published to a third party, whether the defendant is responsible for the publication, and whether a defence applies.

A defendant cannot automatically rely on the responsible-communication defence merely because an AI system generated the statement. The applicable legal requirements must be satisfied.

Limitation: Grant did not concern AI-generated statements. It establishes general Canadian defamation principles that may apply to AI-assisted publication.

Case 6: Crookes v. Newton (2011)

Supreme Court of Canada · 2011 SCC 47 · [2011] 3 SCR 269

Facts: The dispute concerned hyperlinks on a website that directed readers to other online material alleged to be defamatory.

Legal issue: Does providing a hyperlink to defamatory material, without more, amount to publication of that material for purposes of defamation law?

Decision: The Supreme Court held that a hyperlink, by itself, does not constitute publication of the content to which it points. The Court distinguished simply referring readers to material from actually communicating or adopting its defamatory content.

Legal principles:

Online publication must be assessed according to the nature of the communication.

Merely facilitating access to external content is not always equivalent to publishing that content.

Context and the defendant's actual communication remain important.

Application to AI civil claims: An AI platform that retrieves, summarises, quotes, or links to external material may raise difficult questions about whether it has published a defamatory statement. The answer depends on the system's actual output and the applicable publication principles.

A chatbot that independently generates and communicates a false accusation may present a different situation from a platform that merely supplies a hyperlink.

Limitation: Crookes concerns hyperlinks and defamation, not the legal status of AI-generated text.

Case 7: Google Inc. v. Equustek Solutions Inc. (2017)

Supreme Court of Canada · 2017 SCC 34 · [2017] 1 SCR 824

Facts: Equustek Solutions alleged that a former distributor was selling competing products while misusing confidential information and intellectual property. Google was not the principal wrongdoer, but its search engine helped users locate the competing products. Equustek sought an injunction requiring Google to remove the relevant search results globally.

Legal issue: Could a court grant an injunction affecting an online intermediary that was not the primary defendant in the underlying dispute?

Decision: The Supreme Court upheld the interlocutory injunction requiring Google to de-index the specified websites worldwide, subject to the circumstances of the case and the court's jurisdiction.

Legal principles:

Courts may have equitable powers to issue effective injunctions in appropriate cases.

The involvement of an intermediary does not automatically prevent the court from granting relief.

The internet's cross-border character does not necessarily make effective judicial remedies impossible.

Application to AI civil claims: This decision may be relevant when a claimant seeks an injunction against the continued distribution of AI-generated defamatory material, unlawful reproductions, or content alleged to infringe legal rights. A court would still need to assess jurisdiction, the legal basis for the injunction, the rights of the parties, and the appropriate scope of relief.

Limitation: The case concerned search-engine de-indexing and intellectual-property-related disputes, not generative AI. It does not establish automatic liability for platforms hosting or distributing AI content.

Case 8: A.B. v. Bragg Communications Inc. (2012)

Supreme Court of Canada · 2012 SCC 46 · [2012] 2 SCR 567

Facts: A young person sought to identify the individual responsible for creating an allegedly defamatory fake social-media profile using her image and personal information. She also sought permission to proceed anonymously to protect herself from further harm.

Legal issue: How should courts balance open-court principles against privacy and the protection of vulnerable individuals in online defamation proceedings?

Decision: The Supreme Court allowed the young person's request to proceed anonymously in the circumstances. It recognised that online abuse and the particular vulnerability of a minor could justify protecting her identity.

Legal principles:

Open courts are fundamental, but anonymity may be justified in appropriate circumstances.

Courts can take account of the nature of online abuse and the risk of further harm.

Privacy and dignity may be relevant when deciding how proceedings should be conducted.

Application to AI civil claims: An AI-generated fake profile, synthetic image, or fabricated allegation may expose a victim to humiliation, harassment, or further dissemination of personal information. A.B. v. Bragg may inform requests for anonymity or protective procedural measures, particularly where a claimant is a minor or otherwise vulnerable.

Limitation: This case did not decide whether an AI developer or platform was liable for creating the harmful material. Its importance lies in procedural protection and online privacy.

5. What the eight cases establish for Canadian AI accountability

These decisions do not create a single AI-liability test. They show how existing Canadian doctrines can apply to particular forms of AI-related harm.

6. Negligence claims against AI developers and deployers

Negligence is likely to be important in future Canadian AI litigation, particularly where a system causes physical injury, financial loss, or foreseeable harm to an identifiable person.

A claimant will generally need to establish the following elements.

A. Duty of care

The claimant must establish that the defendant owed them a duty of care recognised by law.

Potential defendants include AI developers, employers, professional service providers, hospitals, financial institutions, and businesses deploying AI tools.

The existence and scope of a duty will depend on the relationship between the parties, the foreseeability of harm, and applicable legal principles. A developer does not automatically owe every person affected by a model a duty of care.

B. Breach of the standard of care

The claimant must identify what the defendant did wrong.

Examples may include:

Failing to test a high-risk system adequately.

Deploying a system despite known and material limitations.

Failing to provide appropriate warnings.

Ignoring reasonably foreseeable discriminatory outcomes.

Failing to implement appropriate human oversight.

Allowing sensitive personal information to be exposed through inadequate safeguards.

A court would assess the conduct against the standard required by the circumstances, not simply against an abstract expectation that AI must be perfect.

C. Causation

The claimant must establish the required connection between the defendant's breach and the alleged injury.

For example, a borrower denied a loan by an AI-based scoring system may need to establish that the defendant's legally actionable conduct caused the relevant loss. Evidence may include system logs, decision records, alternative assessments, expert testimony, and the lender's own policies.

Causation may be especially difficult where multiple organisations contributed to the system or where the outcome involved several independent decision-makers.

D. Legally recognised damage

Depending on the cause of action, a claimant may seek compensation for financial loss, physical injury, reputational harm, or another recognised form of injury.

Not every inaccurate AI output creates a damages claim. The claimant must establish the kind of harm and legal elements required by the particular cause of action.

E. Defences and allocation of responsibility

A defendant may argue that the harm resulted from an unforeseeable misuse, an intervening act, the claimant's own conduct, or another party's independent decision.

Contractual allocation of responsibility, contribution claims between defendants, and statutory limitations may also affect the outcome. Contract terms cannot necessarily exclude liability where legislation or public policy prevents that result.

7. AI discrimination and Canadian human rights claims

AI systems can reproduce or amplify discriminatory patterns when their training data, design, variables, or deployment practices disadvantage protected groups.

Examples include:

Recruitment algorithms that systematically screen out applicants based on protected characteristics.

Automated credit assessments that create unjustified barriers to financial services.

Facial-recognition systems that perform differently across demographic groups.

Automated eligibility decisions that disadvantage people with disabilities.

AI-assisted workplace monitoring that produces discriminatory treatment.

Under Canadian human rights legislation, the key issue is often whether the conduct falls within a protected area and adversely affects a person on a prohibited ground.

In employment, for example, a claimant may argue that an AI screening system has a discriminatory effect based on disability, race, sex, age, or another protected ground recognised by the applicable law.

The employer may need to justify its practices under the relevant legal test. The use of an external AI vendor does not automatically excuse an employer from its own human rights obligations.

A claimant may seek remedies available under the relevant human rights statute, potentially including compensation, reconsideration of a decision, changes to employment practices, or other orders within the tribunal's jurisdiction.

Important distinction: A discriminatory AI outcome does not automatically establish negligence, and a negligence claim does not automatically establish discrimination. The legal elements and available remedies differ.

8. Privacy, personal information, and AI accountability

AI systems may collect, infer, retain, and disclose personal information. Civil claims can arise when these activities violate applicable legal obligations.

A. Unauthorised collection or disclosure

A business may face legal scrutiny if it uses customer records, employee information, medical data, or confidential documents to train or operate an AI system without an adequate legal basis.

The relevant question is whether the processing complies with applicable privacy legislation, consent requirements, contractual obligations, and other legal duties.

B. Automated profiling

AI tools may infer sensitive attributes from seemingly ordinary information. For example, an algorithm may infer health conditions, financial vulnerability, or personal preferences from browsing activity or transaction records.

The legal implications depend on what information is collected, how it is used, whether the processing is authorised, and the applicable statutory safeguards.

C. Data breaches and security failures

If an AI service exposes personal information because of inadequate safeguards, potential legal consequences may include regulatory investigations, statutory obligations, contractual claims, and civil litigation where a recognised cause of action is available.

A data breach does not necessarily entitle every affected person to damages. The claimant must establish the relevant legal basis and satisfy its requirements.

D. Quebec privacy law

Quebec's private-sector privacy legislation, as amended by Law 25, includes provisions addressing transparency, consent, and certain decisions based exclusively on automated processing of personal information.

In relevant circumstances, individuals may have rights to be informed about such processing and to request information about the personal information used, the principal factors and parameters involved, and the right to submit observations to a person who can review the decision.

The precise obligations depend on the statute, the type of decision, and the organisation's activities.

9. Defamation and AI-generated misinformation

AI systems can generate false statements about real people, including fabricated criminal allegations, false professional histories, or inaccurate claims about business conduct.

A Canadian defamation claim generally requires consideration of three core elements:

The words were defamatory in the sense that they would tend to lower the claimant's reputation in the eyes of a reasonable person.

The words referred to the claimant.

The words were communicated to at least one person other than the claimant.

Defences may include justification, fair comment, privilege, and responsible communication on matters of public interest, depending on the circumstances.

Who may be responsible?

Potential defendants could include a publisher, an organisation that distributed AI-generated content, or another party that legally participated in publication. Whether an AI developer or platform is liable depends on its conduct, the relevant legal principles, and the facts.

Questions may include whether the defendant generated the statement, knowingly adopted it, republished it, or exercised control over the relevant communication.

The fact that a statement was generated by an algorithm is not itself a complete defence. Equally, the involvement of an AI tool does not automatically make its developer liable for every output.

10. Contractual claims and commercial AI disputes

AI accountability can also arise from commercial contracts.

Suppose a Canadian company pays a supplier for an AI system represented as capable of accurately detecting fraudulent transactions. The system repeatedly generates false positives, causing substantial financial loss.

The dispute may involve:

Whether the supplier made an enforceable contractual promise.

Whether the system met agreed performance requirements.

Whether the supplier breached express or implied contractual terms.

Whether the customer relied on a misleading representation.

Whether liability limitations or exclusions apply.

Whether the losses were caused by the breach and are recoverable under the contract.

In Quebec, Article 1458 of the Civil Code of Québec is relevant to contractual civil liability. In common-law provinces, contractual interpretation, breach, causation, remoteness, and damages principles will generally be central.

AI contracts should therefore specify performance standards, testing obligations, security requirements, disclosure duties, audit rights, incident reporting, and allocation of responsibility for system failures.

11. Remedies available in Canadian AI civil claims

The appropriate remedy depends on the cause of action, the jurisdiction, and the evidence.

Compensatory damages

Compensation for legally recognised losses, such as financial loss, personal injury, or reputational harm, where the claimant establishes entitlement.

Injunctions

Court orders restricting or requiring conduct, potentially including the removal of unlawful content or cessation of a harmful practice where the legal requirements are met.

Disclosure and evidence preservation

Procedural orders or litigation steps to obtain relevant records, decision logs, contracts, and other evidence, subject to privilege, privacy, proportionality, and procedural rules.

Class proceedings

Collective litigation may be available where numerous people have sufficiently related claims and the applicable certification requirements are satisfied.

Human rights and regulatory remedies

Depending on the governing legislation, remedies may include compensation, policy changes, reconsideration of decisions, compliance orders, or regulatory penalties.

A claimant should not assume that every remedy is available in the same proceeding. Regulatory complaints, civil damages actions, and human rights proceedings have different procedures, jurisdictions, and requirements.

12. Practical steps for bringing an AI-related civil claim

A person considering litigation should ordinarily:

Identify the harm. Specify whether the issue involves financial loss, discrimination, privacy, defamation, contractual breach, or physical injury.

Identify responsible parties. Determine who developed, supplied, selected, deployed, or relied on the AI system.

Preserve evidence. Retain screenshots, AI outputs, correspondence, contracts, decision notices, relevant records, and information identifying when the output was generated.

Document causation and loss. Record the consequences of the decision or statement and preserve relevant financial, medical, professional, or reputational evidence.

Determine the governing law. Identify the province, applicable legislation, contractual terms, and any cross-border jurisdiction issues.

Consider pre-litigation remedies. Depending on the case, these may include an internal review, privacy complaint, human rights complaint, demand letter, or request for correction.

Check limitation periods. Time limits differ by claim and jurisdiction. Some statutory or specialised proceedings have shorter deadlines than ordinary civil claims.

Assess expert evidence. Complex cases may require technical evidence about model behaviour, data provenance, testing, system logs, or the causal relationship between an algorithm and the loss.

An injured person should avoid assuming that an AI system's internal operation must be fully understood before a claim can be investigated. However, establishing the relevant legal elements may require disclosure and expert evidence.

13. Challenges in Canadian AI civil litigation

Several recurring challenges make these claims difficult.

Attribution of responsibility: An AI outcome may reflect the combined contributions of a model developer, data supplier, software integrator, employer, and end user.

Proof of causation: It can be difficult to establish how a particular output was produced and whether the defendant's conduct caused the loss.

Access to evidence: Important information may be held in proprietary systems or internal records that are not immediately available to a claimant.

Bias and discrimination: Demonstrating a discriminatory effect may require statistical analysis, comparator evidence, or technical expertise.

Jurisdiction: An AI provider may be located outside Canada, raising questions about jurisdiction, service, applicable law, and enforcement.

Contractual restrictions: Standard-form terms may contain liability limitations, arbitration clauses, or foreign forum-selection provisions. Their enforceability must be examined under the applicable law.

Rapid technological change: Existing doctrines must be applied to new systems without assuming that every AI-specific risk already has a settled legal rule.

14. Conclusion

Canadian civil law can provide remedies for certain harms caused by AI systems, but liability depends on the applicable cause of action and the evidence.

The eight decisions examined above contribute different parts of the legal framework:

Moffatt v. Air Canada directly addresses misleading information provided by a chatbot.

Douez v. Facebook examines privacy rights and online contractual terms.

R. v. Bykovets addresses privacy in digital identifiers.

Jones v. Tsige recognises intrusion upon seclusion in Ontario.

Grant v. Torstar establishes principles governing defamation and responsible communication.

Crookes v. Newton addresses online publication and hyperlinks.

Google v. Equustek concerns injunctive relief involving an online intermediary.

A.B. v. Bragg addresses privacy and anonymity in online proceedings.

The broader lesson is that AI is not a substitute for legal analysis of responsibility. Courts must identify the conduct of the relevant parties, apply the correct legal standard, determine causation and legally recognised harm, and select a remedy authorised by law.

Canada's existing civil-law principles provide a foundation for AI accountability, but many questions involving generative AI, automated decision-making, and foundation models remain fact-dependent or legally unsettled. The cited decisions should therefore be distinguished carefully between direct AI authority and decisions whose principles may be applied by analogy.

Educational note: This explanation is intended for legal study and general information. A real claim requires analysis of the province or territory involved, the applicable statute, the precise harm, the defendant's role, and the relevant limitation period.