A Model AI Practice Direction for Canadian Courts and Tribunals

Two-thirds of the Canadian courts and tribunals we track have no artificial intelligence policy at all. This is a model AI practice direction any court or tribunal can adopt, assembled entirely from rules already in force across Canada.

A model AI practice direction for Canadian courts and tribunals

This template accompanies a Courtready.ca study, “Can I Use AI in Court? Artificial Intelligence Policies at 249 Canadian Courts and Tribunals“, which found that two-thirds of Canadian adjudicative bodies tracked have no AI policy at all. Of the 82 that do, the author reviewed the 45 distinct policy documents behind them. That review suggests a sound policy should do three things: it requires parties to disclose their use of AI, without penalising those who do; it teaches parties how to catch AI’s errors rather than simply warning them to be careful; and it gives parties a way to correct a filing once they discover an AI-related mistake. Each of these elements already exists in a Canadian policy in force today, but no single policy combines all three.

167 of 249 Canadian courts and tribunals we tracked have no AI policy at all. This template is for them.

The Model AI Practice Direction

Practice Direction: Using Artificial Intelligence in Proceedings Before [NAME OF BODY]

Effective: [DATE]. Applies to all participants (parties, lawyers, representatives, witnesses and interveners), whether or not they have a lawyer.

1. What we mean by AI

Tools that create new content, such as text, images, audio or video, in response to what you type. Examples: ChatGPT, Google Gemini, Microsoft Copilot, Claude.1 This Direction does not apply to tools that only assist with mechanics: spell-check, grammar-check, formatting, dictation or speech-to-text.2 Be aware that ordinary search engines now show AI-written answers at the top of the results page.3

2. What you may use AI for

Understanding legal words, forms or procedures; organizing your thoughts; summarizing documents you already have; drafting or improving your writing; translating; simplifying language; finding starting points for research, which you must then verify.4 Using AI is not, by itself, a problem, and we will not think less of your case because you used it.5

3. Protect privacy

Whatever you type into a public AI tool may be stored, reused, and seen by others, and you usually cannot get it back. Do not enter personal, financial, or confidential information into a public AI tool.6 Do not enter any part of the record or any document you received because of this proceeding.7

4. What AI must never be used for

(a) Creating witness statements, affidavits, or anything sworn or affirmed. These must come from a real person’s own knowledge and experience.8

(b) Creating or altering photographs, video, audio, documents or messages intended as evidence.9

(c) Citing cases, laws or quotations you have not confirmed are real.10

(d) Recording, transcribing or attending a hearing (including AI note-taking bots), without our prior permission.11

5. Check everything before you file it

AI can invent court cases that do not exist; it can also invent quotations from real cases that are not present in those cases.12 Before relying on any authority an AI tool gave you: find it on CanLII (canlii.org) or on our website; if you cannot find it, assume it does not exist and do not use it;13 read it to confirm it says what you were told it says;14 and cite it with a link to a free public source, or attach a copy.15 Do not rely on AI-written case summaries; read the decision.16

You are responsible for everything you file and everything you say at a hearing, even if AI helped prepare it.17 Appendix B sets out the steps in plain language.

6. Tell us when you used AI to generate content

6.1 If AI created or generated content in a document you file, say so in the document’s first paragraph, using the form in Appendix A.18

6.2 No declaration is needed if you only used AI to: check spelling or grammar; format; dictate; suggest changes to writing you had already created yourself;19 [translate a document into or out of a language you understand; put a document into simpler language so you can understand it; or assist you because of a disability, a literacy barrier, or a language barrier.]20

6.3 The declaration, in and of itself, will not attract any adverse inference.21

7. If you discover an error, correct it

If you learn that something you filed contains a case that does not exist, an invented quotation, or inaccurate information, notify the other parties and us as soon as possible and correct or withdraw it.22

8. Consequences

Our response will depend on the facts, including whether you had a lawyer, whether the error was honest, and whether you corrected it yourself.23 An honest mistake by an unrepresented person, corrected when discovered, may be addressed by simply not relying on the material.24 Serious responses, such as striking material, costs, dismissal, or regulatory referral, are reserved for deliberate conduct: defending invented authorities, fabricating evidence, or lying about AI use.

Appendix A: Declaration form

Copy this into the first paragraph of your document. Fill in the blanks. Delete whichever option does not apply.

DECLARATION REGARDING THE USE OF ARTIFICIAL INTELLIGENCE

Artificial intelligence was used to generate content in this document:

☐ throughout the document    ☐ at paragraphs ______ to ______

I confirm that: (1) I have read all of the content that AI generated; (2) I have checked that every case, statute, regulation and quotation in this document is real and says what I have said it says; and (3) I did not use AI to create or alter any evidence.

Name: ____________________   ☐ Party   ☐ Lawyer   ☐ Representative   ☐ Self-represented litigant

Date: ____________   Signature: ____________________

Appendix B: How to check whether a case is real

AI tools invent court cases. Here is how to catch them in about two minutes.

  • Go to CanLII. Open canlii.org. It is free. You do not need an account.
  • Search the case name. Type the party names into the search box.
  • If nothing comes up, the case might be fake. Do not assume you searched wrong. Try once more with the citation (the string of numbers and letters). If that also finds nothing, take it out of the document.
  • If it does come up, open it and read it. AI often takes a real case and tells you it decided something it did not decide. If you cannot find that idea anywhere in the decision, the AI made it up, even though the case is real.
  • Check any quotation. Use Ctrl+F (or Command+F) and search for a few of the quoted words inside the decision. If the words are not there, the quotation is invented. Take it out.
  • Copy the link. Put the CanLII web address in your document next to the case name, with the paragraph number you rely on.
  • A trap to avoid: some websites show AI-written summaries of cases. Do not rely on the summary. Read the decision.

Notes and sources

  1. Definitions adapted from Canada Industrial Relations Board, Policy on the Use of Generative Artificial Intelligence by Parties (effective Nov. 1, 2025), online: https://www.cirb-ccri.gc.ca/en/resources/policy-use-generative-artificial [CIRB Policy]. ↩
  2. Federal Court, Notice to the Parties and the Profession: The Use of Artificial Intelligence in Court Proceedings (Dec. 20, 2023, updated May 7, 2024), online: https://www.fct-cf.gc.ca/Content/assets/pdf/base/FC-Updated-AI-Notice-EN.pdf [FC Notice] (the Notice “does not apply to AI that lacks the creative ability to generate new content”). ↩
  3. Social Security Tribunal of Canada, Using Artificial Intelligence in Appeals at the Social Security Tribunal (July 9, 2026), s. 1.3, online: https://www.sst-tss.gc.ca/en/decisions-laws-rules-and-policies/using-artificial-intelligence-appeals-social-security-tribunal [SST Policy]. This is the only Canadian policy reviewed to warn of this. ↩
  4. BC Employment Standards Tribunal, Practice Directive 2026-002: Use of Artificial Intelligence (July 1, 2026), s. 4.1 (drafting, organizing, summarizing, translating, formatting), online: https://www.bcest.bc.ca/pd-2026-002-use-of-ai-2026-07-01/ [BCEST Practice Directive]; Court of King’s Bench for Saskatchewan, General Application Practice Directive #12: Use of Artificial Intelligence in Court Submissions (effective Jan. 1, 2026), s. 3 (“Nothing in this Practice Directive prohibits the appropriate use of AI”), online: https://sasklawcourts.ca/wp-content/uploads/2026/05/KB_GA-PD-12.pdf [Saskatchewan KB Directive]. ↩
  5. Canadian International Trade Tribunal, Use of Artificial Intelligence (AI) (Jan. 30, 2026) (“participants will not be penalized simply for using AI tools”), online: https://www.citt-tcce.gc.ca/en/practices-and-procedures/use-artificial-intelligence-ai [CITT Practice Notice]. ↩
  6. CITT Practice Notice (prohibition on inputting third-party confidential information); see also the privacy warnings in Condominium Authority Tribunal, Practice Direction: Use of Artificial Intelligence in CAT Cases (effective Dec. 1, 2024), online: https://www.condoauthorityontario.ca/wp-content/uploads/2024/11/Practice-Direction-AI-use-in-Tribunal-Proceedings.pdf [CAT Practice Direction], and the CIRB Policy. ↩
  7. BC Health Professions Review Board, Practice Directive 9: Use of Artificial Intelligence (AI) in Tribunal Proceedings (Jan. 27, 2026) (“Parties are not permitted to put any part of the Record into an AI tool such as ChatGPT”), online: https://www.bchprb.ca/app/uploads/sites/791/2026/01/Practice-Directive-9-Use-of-Artificial-Intelligence.pdf [HPRB Practice Directive]. ↩
  8. CIRB Policy; Manitoba Labour Board, Practice Directive on the Use of Generative Artificial Intelligence (“It is critical that these be based on a person’s own knowledge and experience”), online: https://www.manitobalabourboard.ca/resources-forms/practice-directives.html [MLB Practice Directive]. ↩
  9. BC Civil Resolution Tribunal, Standard Rules (effective Feb. 1, 2026), r. 8.3(9), online: https://civilresolutionbc.ca/wp-content/uploads/CRT-Standard-Rules-effective-February-1-2026.pdf [BCCRT Standard Rules]; BCEST Practice Directive, s. 6.1. ↩
  10. BCCRT Standard Rules, r. 8.3(8) (“A party must not include in their arguments nonexistent cases or legislation, such as those created by an artificial intelligence tool”). ↩
  11. BCEST Practice Directive, s. 6.2. ↩
  12. Tom Macintosh Zheng, AI Hallucinated Cases in Canadian Courts: Database of Fictitious Citations (Toronto: Courtready, 2026), online: https://courtready.ca/fictitious-citations-in-canadian-courts/. ↩
  13. SST Policy, s. 3.2 (“If you can’t find the case, it may not exist”). ↩
  14. SST Policy, s. 3.2; CAT Practice Direction, s. 5 (“Read the decision to make sure it says what you think it is supposed to say”). ↩
  15. Federal Public Sector Labour Relations and Employment Board, Practice Direction: Use of Artificial Intelligence (AI) in Board Proceedings (requiring a citation and hyperlink, or an attached copy), online: https://pslreb-crtefp.gc.ca/en/resources/policies/artificial-intelligence.html [FPSLREB Practice Direction]; SST Policy, ss. 3.3 to 3.4; MLB Practice Directive. ↩
  16. CIRB Policy (parties may rely on “CanLII (excluding AI generated summaries)”). ↩
  17. Supreme Court of Yukon, Practice Direction GENERAL-35: Conditions on Use of Generative AI in Written and Oral Representations (June 2, 2026), online: https://www.yukoncourts.ca/sites/default/files/2026-06/35.%20General%2035%20-Conditions%20on%20use%20of%20generative%20AI%20in%20written%20and%20oral%20representations_1.pdf [Yukon GENERAL-35]. The responsibility extends to written and oral representations. ↩
  18. The first-paragraph declaration model originated in the FC Notice and was adopted in: Ontario Land Tribunal, Practice Direction on the Use of Artificial Intelligence (effective Mar. 30, 2026), online: https://olt.gov.on.ca/wp-content/uploads/AI-Practice-Direction.html [OLT Practice Direction]; Trademarks Opposition Board, Use of AI in Proceedings Before the Trademarks Opposition Board (June 4, 2025), online: https://ised-isde.canada.ca/site/canadian-intellectual-property-office/en/trademarks-opposition-board/use-ai-proceedings-trademarks-opposition-board [TMOB Practice Notice]; and Supreme Court of the Northwest Territories, Notice to the Public and Practice Direction: The Use of Artificial Intelligence in Court Proceedings (Oct. 2025), online: https://www.nwtcourts.ca/en/files/notices-and-directives/sc/Use%20of%20Artificial%20Intelligence%20in%20Court%20Proceeding.October%2029%2C%202025.pdf [NWT Notice]. ↩
  19. FC Notice (“a Declaration is not required if AI was used to merely suggest changes, provide recommendations, or critique content already created by a human”); BCEST Practice Directive, s. 5.3. ↩
  20. The exemptions in bold have no precedent in any current Canadian policy reviewed. In the author’s view, requiring a person to announce that they needed help reading their own court file is a barrier. But if AI wrote the substance of your argument, or found your cases, declare it. ↩
  21. FC Notice, “Neutrality” (“the inclusion of a Declaration, in and of itself, will not attract an adverse inference”); OLT Practice Direction (same assurance). Of the nineteen Canadian bodies requiring disclosure, only these two attach this assurance to the declaration. ↩
  22. BC Property Assessment Appeal Board, Participant Code of Conduct (current as of Apr. 2, 2025), s. 6.2, online: https://www.assessmentappeal.bc.ca/download_file/view/407/ [PAAB Code of Conduct]. This is the only party-side duty to correct among the 45 policies reviewed. ↩
  23. Ontario Superior Court of Justice, Consolidated Civil Provincial Practice Direction, Part 12 (Use of Artificial Intelligence), para. 128 (“the court’s response will depend on the specific facts and circumstances of the case”), online: https://www.ontariocourts.ca/scj/filing-procedures/provincial/consolidated-civil-provincial-practice-direction/. ↩
  24. SST Policy, s. 4.3 (a member “may place less weight” on AI-generated material containing errors). ↩

Every element of a good AI policy already exists in Canada.

Methodology & Sources

This model practice direction is assembled from 45 artificial intelligence policies in force across 249 Canadian courts and tribunals, reviewed in July 2026. Every provision is footnoted to the Canadian court or tribunal policy it draws on. The three genuinely novel provisions (the accessibility and translation exemptions in section 6.2) are marked in the notes. Full study and sources at courtready.ca.