AI in Canadian Law: What Courts, Law Societies and Privacy Law Require
By BrainBridge, makers of BrainLex. Last updated . General information, not legal advice.
Canadian lawyers can use generative AI, but courts, law societies and privacy law all expect the same things: understand the tool, check every authority it produces, protect client information and follow each court's rules on disclosure. The risk is not hypothetical. As of October 13, 2026, the Courtready database lists 267 Canadian decisions in 64 courts and tribunals that flagged AI hallucinations or other AI misuse since January 2024.
Key Points
- Verify every authority. In Ontario, every factum must include a signed statement that each cited authority is authentic (Rules of Civil Procedure, r. 4.06.1(2.1), in force December 1, 2024).
- Disclose where required. The Federal Court requires a declaration in the first paragraph of any filed document containing AI-generated content. Courts in Manitoba, Yukon and the Nova Scotia Provincial Court also require parties to explain AI use.
- Responsibility stays with counsel. The Court of Appeal of Alberta held lead counsel responsible for a contractor's AI-drafted factum and later ordered costs against him personally (Reddy v Saroya).
- Law societies have published guidance. At least nine Canadian law societies, from British Columbia to Newfoundland and Labrador, have issued guidance on generative AI.
- Privacy duties follow the data. Under PIPEDA, a firm stays accountable for personal information it sends to an AI vendor for processing, including a vendor outside Canada.
How Often Canadian Courts See AI Hallucinations
The Courtready database of Canadian decisions (updated October 13, 2026) counts 267 decisions in 64 courts and tribunals that identified AI hallucinations or other AI misuse since January 1, 2024. Most involve fictitious citations (221 decisions), and 83% involve a self-represented party. Of the 263 decisions with a recorded sanction outcome, 60 imposed a serious sanction, 111 issued a warning and 92 recorded no sanction.
Lawyers appear in a minority of these decisions, but the cases involving counsel set the standard for the profession. Five decisions show how Canadian courts respond:
Zhang v. Chen, 2024 BCSC 285
Supreme Court of British Columbia
Counsel cited two non-existent cases generated by ChatGPT. The court ordered counsel to personally bear costs for the extra work the fake citations caused.
Ko v. Li, 2025 ONSC 2766
Ontario Superior Court of Justice
A factum cited cases that did not exist or did not support the propositions cited. The court ordered counsel to show cause why she should not be found in contempt. The proceeding was dismissed after she admitted the error, apologized, corrected the factum and committed to continuing professional development on AI (2025 ONSC 2965).
Hussein v. Canada (Citizenship and Immigration), 2025 FC 1060
Federal Court
Counsel relied on Visto.ai, a legal AI tool built for Canadian immigration practitioners, which produced non-existent cases. The court found that disclosing AI use only after four directions amounted to an attempt to mislead, and costs of $100 were later ordered against counsel personally (2025 FC 1138). A law-specific tool is not safe by default.
Specter Aviation Limited c. Laprade, 2025 QCCS 3521
Superior Court of Quebec
A self-represented party was ordered to pay $5,000 for filing submissions with AI-invented jurisprudence. The court identified eight instances of non-existent citations, decisions not rendered and inconsistent conclusions.
Reddy v Saroya, 2025 ABCA 322
Court of Appeal of Alberta
A contractor drafted a factum that cited seven cases that could not be found. The court held that lead counsel is responsible for everything filed under his name, even though the contractor had assured him AI was not used, and in a later costs decision ordered costs against him personally (2026 ABCA 20).
More detail on these decisions: the Canadian court record on AI hallucinations.
Court Practice Directions on AI
Canadian courts take two approaches. Some require parties to disclose AI use. Others require a human to verify every AI-assisted submission without a disclosure rule. Either way, the filer is responsible for accuracy.
| Court | Issued | What it requires |
|---|---|---|
| Federal Court | Dec 20, 2023; updated May 7, 2024 | A declaration in the first paragraph of any document containing AI-generated content. Not needed when AI only suggests or critiques text a person wrote. |
| Ontario Rules of Civil Procedure and Superior Court of Justice | Dec 1, 2024; Feb 2026 | Every factum must include a signed statement certifying the authenticity of each authority cited. The Superior Court's practice directions address AI use in civil, family and criminal proceedings. |
| Court of King's Bench of Manitoba | June 23, 2023 | Materials prepared with AI must indicate how AI was used. |
| Supreme Court of Yukon | June 26, 2023; June 2, 2026 | Disclose the AI tool used and its purpose. The 2026 direction sets conditions for AI in written and oral submissions, with costs, including personal costs, for non-compliance. |
| Alberta Court of Appeal, Court of King's Bench and Court of Justice | Oct 6, 2023 | Rely on authoritative sources and keep a human in the loop. No disclosure requirement. |
| Court of King's Bench for Saskatchewan | Jan 2026 | Appropriate AI use is allowed. Filers are responsible for the accuracy, authenticity and reliability of what they file. |
| Court of Appeal for British Columbia | Mar 12, 2024 | The registrar's filing directive makes filers responsible for the authenticity and accuracy of their materials. |
| Provincial Court of British Columbia | Plain-language guidance July 2026 | Lawyers and litigants must disclose the extent of their AI use when asked. |
| Quebec Court of Appeal | Aug 8, 2024 | Caution and human involvement. Litigants remain accountable for the accuracy of their submissions. |
| Court of Québec | Jan 26, 2024 | AI-generated material requires serious human review and cross-referencing. |
| Supreme Court of Nova Scotia | Oct 18, 2023; Feb 2025 | Use authoritative sources and verify AI-generated submissions with meaningful human control. |
| Provincial Court of Nova Scotia | Oct 27, 2023 | Explain how AI was used, and source legal references from established legal databases. |
| Nova Scotia Court of Appeal | Mar 2025 | Participants are responsible for accuracy and must source legal references from accredited databases. |
| Supreme Court of Newfoundland and Labrador | Oct 12, 2023 | A human must verify AI-generated submissions. No disclosure requirement. |
| Supreme Court of the Northwest Territories | Oct 2025 | Caution, credible sources and human involvement. |
Judges have their own framework: the Canadian Judicial Council's Guidelines for the Use of Artificial Intelligence in Canadian Courts (2024). Practice directions change often, so check the current version on the court's website before you file.
Law Society Guidance on Generative AI
The guidance mostly applies duties lawyers already have: competence, confidentiality, supervision and candour to the court.
| Law society | Guidance | Date |
|---|---|---|
| Law Society of British Columbia | Guidance on Professional Responsibility and Generative AI | Oct 2023 |
| Law Society of Alberta | The Generative AI Playbook | Jan 2024 |
| Law Society of Saskatchewan | Guidelines for the Use of Generative Artificial Intelligence in the Practice of Law | Feb 2024 |
| Law Society of Manitoba | Generative Artificial Intelligence: Guidelines for Use in the Practice of Law | Apr 2024 |
| Law Society of Ontario | White paper: Licensee use of generative artificial intelligence, available from the LSO Technology Resource Centre | Apr 2024 |
| Barreau du Québec | Practical guide to the responsible use of generative AI (French only) | Oct 2024 |
| Nova Scotia Barristers' Society | Artificial Intelligence in the Practice of Law (short guide) | Dec 2023 |
| Law Society of Newfoundland and Labrador | Artificial Intelligence in Your Practice | Undated |
| Law Society of the Northwest Territories | Guidelines for the Use of Generative AI in the Practice of Law | Jan 2025 |
What the Guidance Has in Common
- Competence: understand how the tool works and where it fails, and keep learning as it changes.
- Verification: check every output, including citations, against reliable sources such as CanLII, official court websites or commercial publishers.
- Confidentiality: do not put client information into a tool without safeguards that protect it.
- Supervision: set firm policies on AI and review AI-assisted work by junior lawyers and staff.
- Candour: never put fabricated authorities before a court, and follow each court's disclosure rules.
- Fees: bill for the work you did, such as prompting, reviewing and editing, not for time the tool saved.
- Client communication: consider telling clients when AI is used on their matters, especially when their data is involved.
Summary based on the Law Society of Alberta's Gen AI Rules of Engagement for Canadian Lawyers.
Privacy, Confidentiality and Where Your Data Goes
PIPEDA makes you accountable for your vendors. Under Principle 4.1.3 of the Personal Information Protection and Electronic Documents Act, an organization remains responsible for personal information it transfers to a third party for processing, and must use contractual or other means to provide a comparable level of protection. That includes an AI vendor processing data outside Canada.
Quebec requires an assessment first. Since September 22, 2023, section 17 of Quebec's Act respecting the protection of personal information in the private sector requires a privacy impact assessment before personal information is communicated outside Quebec. The transfer can proceed only if the information would receive adequate protection.
Professional confidentiality is broader than privacy law. Law society codes of conduct, at rule 3.3 in most provinces, require lawyers to hold all information about a client's affairs in strict confidence, whether or not it is personal information.
Federal reform is pending. Bill C-36, introduced on June 15, 2026, would enact the Protecting Privacy and Consumer Data Act and amend PIPEDA. As introduced, it provides for administrative penalties of up to the greater of $10 million or 3% of gross global revenue, and fines of up to the greater of $25 million or 5% for the most serious offences. As of October 2026 it is at second reading in the House of Commons, so PIPEDA still applies.
Checklist: Questions to Ask a Legal AI Vendor in Canada
- Where is our data processed and stored, and which country's laws apply to the vendor?
- Is our data used to train or improve the vendor's models, and can the contract rule that out?
- How does the tool check that a cited case exists and says what it claims, and against which sources, such as CanLII?
- Can we export an audit trail of prompts, outputs, sources and verification results for a matter?
- Does it help us meet court disclosure rules, such as the Federal Court's AI declaration?
- Does it cover Canadian law across provinces, including Quebec civil law and French-language sources?
- Can it be deployed privately, in a dedicated cloud environment or on our own hardware?
- Which security controls does it have: encryption, single sign-on, role-based access and independent audits such as SOC 2 or ISO 27001?
- Is it priced per seat, per query or at a flat rate?
- What happens to our data when we leave?
How BrainLex Approaches These Requirements
BrainLex is the private legal AI system we are building for Canadian law firms around these obligations.
- Canadian data sovereignty: data is processed and stored only in AWS Canada Central (ca-central-1).
- Dual-model validation: every output is cross-validated by a second AI model before it reaches you.
- Full audit trail: every query, output and validation result is logged with its sources.
- Private deployment: a managed private cloud or on-premise hardware.
- Flat monthly pricing: no per-query charges.
Frequently Asked Questions
Can lawyers in Canada use ChatGPT and other generative AI tools?
Yes. The law society guidance published so far permits generative AI, provided lawyers understand the tool, protect client confidentiality, supervise its use and verify everything it produces. Courts hold counsel responsible for anything filed under their name, whoever or whatever drafted it.
Do I have to disclose AI use to a Canadian court?
It depends on the court. The Federal Court requires a declaration in the first paragraph of any document containing AI-generated content, and the Court of King's Bench of Manitoba, the Supreme Court of Yukon and the Nova Scotia Provincial Court require parties to explain how AI was used. Other courts, such as Alberta's, focus on human verification instead. Check the current practice direction of the court you are filing in.
What happens if a lawyer files AI-generated fake case law in Canada?
Canadian courts have ordered costs against counsel personally (Zhang v. Chen, Hussein v. Canada, Reddy v Saroya) and started contempt proceedings (Ko v. Li). Of 263 Canadian decisions with a recorded sanction outcome in the Courtready database as of October 13, 2026, 60 imposed a serious sanction.
Can a Canadian law firm use a US-hosted AI tool?
PIPEDA does not prohibit transferring personal information outside Canada, but the firm remains accountable for it and must use contractual or other means to provide a comparable level of protection. Quebec requires a privacy impact assessment before personal information is communicated outside Quebec, and every lawyer's professional duty of confidentiality still applies. Choosing a tool that processes data in Canada reduces this exposure.
Which Canadian law societies have published generative AI guidance?
At least nine: the Law Society of British Columbia, the Law Society of Alberta, the Law Society of Saskatchewan, the Law Society of Manitoba, the Law Society of Ontario, the Barreau du Québec, the Nova Scotia Barristers' Society, the Law Society of Newfoundland and Labrador and the Law Society of the Northwest Territories.
Sources
- Courtready, Fictitious Citations in Canadian Courts database, updated October 13, 2026.
- Federal Court, Notice to the Parties and the Profession: The Use of Artificial Intelligence in Court Proceedings, updated May 7, 2024.
- Ontario, O. Reg. 384/24 amending the Rules of Civil Procedure, in force December 1, 2024.
- Law Society of Alberta, Gen AI Rules of Engagement for Canadian Lawyers, updated March 2025.
- Canadian Lawyer, How Canada's courts are regulating, using, and evaluating generative AI, June 10, 2026, updated September 2026.
- Bennett Jones, Requirements and Guidelines From Canadian Regulators, Public Bodies and Courts (on Hussein v. Canada, 2025 FC 1138).
- ABlawg, The Alberta Court of Appeal Weighs In on the Use of AI in Court Submissions, October 15, 2025.
- PIPEDA, Schedule 1, Principle 4.1.3.
- Quebec, Act respecting the protection of personal information in the private sector, CQLR c P-39.1, s. 17.
- Parliament of Canada, LEGISinfo: Bill C-36.
This guide is general information about Canadian law as of October 13, 2026, and is not legal advice. Court and law society requirements change often. Check the current version of each source before relying on it.