Treadstone Associates
Article · Court rules

Do Canadian courts require AI disclosure?

There is no single national rule. The Federal Court requires a declaration when AI generated content in a document prepared for litigation; other courts have issued cautions rather than disclosure requirements.

Treadstone Associates · Updated 2026

Key takeaways

  • • The Federal Court requires a Declaration in the first paragraph of a document prepared for litigation where it contains content created or generated by AI — see the Notice to the Parties and the Profession dated 7 May 2024.
  • • The trigger is co-authorship. A declaration is not required where AI merely suggested changes, made recommendations, or critiqued content already created by a human.
  • • Certified Tribunal Records and expert reports are outside the Notice; expert reports are dealt with through the methodology summary under the Expert Witnesses Code of Conduct.
  • • A declaration attracts no adverse inference. The Court says so expressly, and the party signing still bears responsibility for accuracy either way.

The short answer

Some Canadian courts require disclosure and some do not, so the honest answer is that you check the court you are in. The clearest and most detailed requirement is the Federal Court's. Its Notice to the Parties and the Profession on the Use of Artificial Intelligence in Court Proceedings, dated 7 May 2024, states that the Court expects parties to inform it, and each other, if documents they submit that have been prepared for the purposes of litigation include content created by AI, and that this is to be done by a Declaration in the first paragraph stating that AI was used in preparing the document, either in its entirety or only for specifically identified paragraphs.

The Court supplies the wording. The sample Declaration reads: “Artificial intelligence (AI) was used to generate content in this document at paragraphs 20-30.” The French sample appears alongside it.

What triggers a declaration and what does not

This is the part most summaries get wrong, and the distinction is workable. The Notice says a Declaration is required if content in the material was directly provided by AI, whether or not it was inserted from an external source such as a web-based generative AI. It then says a Declaration is not required if AI was used merely to suggest changes, provide recommendations, or critique content already created by a human who could then consider and manually implement the changes. The Court states the test plainly: a Declaration is required when the role AI plays resembles that of a co-author.

The Notice also draws a boundary around the kind of technology it covers. The requirement applies to AI defined as a computer system capable of generating new content and independently creating or generating information or documents, usually based on prompts. It does not apply to AI that lacks the creative ability to generate new content — the Notice gives system automation, voice recognition and document editing as examples of programs that only follow pre-set instructions.

  • Declare: paragraphs whose text was generated by a model and kept, in a document prepared for litigation and submitted to the Court.
  • Do not declare: a model that critiqued your draft, suggested edits you considered and typed yourself, or transcribed dictation.
  • Outside the Notice: Certified Tribunal Records submitted by tribunals or other third-party decision-makers.
  • Handled elsewhere: expert reports — the Court's understanding is that AI use ought to be disclosed in the summary of methodology under subparagraph 3(i) of the Expert Witnesses Code of Conduct referred to in Rule 52.2 of the Federal Courts Rules.

Two things the Notice makes easier, not harder

First, neutrality. The Court confirms that the inclusion of a Declaration will not in and of itself attract an adverse inference, and that any use of AI which does not generate content within the scope of the Notice will not attract an adverse inference either. The primary purpose of the Declaration is simply to notify the other party and the Court that AI has been used to generate content.

Second, responsibility is unchanged. The Notice states that parties and interveners continue to be held to the existing standards under the Federal Courts Rules, and that the party signing a document submitted to the Court bears responsibility for the accuracy and veracity of its contents. Disclosure is not a transfer of risk, which is a useful thing to say to anyone who thinks declaring is a substitute for checking.

There is also a succession rule that firms miss. Where counsel takes over a matter from a previous lawyer or a self-represented litigant, the Notice says it is reasonable to expect new counsel to make best efforts to ascertain whether previously filed or transferred draft materials contain AI-generated content, and to provide a Declaration for any materials they have reason to believe may include it. If you take carriage of files, that is a question for your intake checklist.

Other courts, and the shape of the difference

Not every court has taken the disclosure route. Alberta's three courts — the Court of Appeal of Alberta, the Court of King's Bench of Alberta and the Alberta Court of Justice — issued a joint notice on large language models, published as NPP 2025-06 on the Court of King's Bench notices to the profession page and dated 6 October 2023. Its emphasis is on reliance and verification rather than on declaring use.

Separately, and often confused with the rule for parties, the Federal Court has published Interim Principles and Guidelines on the Court's own use of Artificial Intelligence, dated 29 September 2025. That policy governs the Court, not you: it commits the Court not to use AI, and specifically automated decision-making tools, in making its judgments and orders without first engaging in public consultation, and sets out principles including accountability, respect of fundamental rights, non-discrimination, accuracy, transparency, cybersecurity and “human in the loop”. Do not cite it as the rule that applies to your factum.

Worked example (illustrative)

A firm in Ontario with a Federal Court immigration practice adopts a single operating rule: any paragraph whose text originated with a model is tagged in the document management system at the moment it is written, by the person who wrote it. At filing, the tags produce the paragraph range for the Declaration automatically.

The reason to tag at drafting rather than reconstruct at filing is that reconstruction is guesswork three weeks later, and the Notice asks for specifically identified paragraphs. The firm's rule for a paragraph that was model-drafted and then substantially rewritten by a lawyer is to keep the tag — declaring more than strictly required costs nothing, because the Court has said a Declaration attracts no adverse inference.

The measurable is the proportion of filed documents where the Declaration was assembled from tags rather than from memory. That is a process metric, and it is the only kind available here.

Where the professional rules sit on top

Court practice directions govern filings. Your law society governs you, and the duties there are broader. LSO rule 3.1-2 requires legal services to be performed to the standard of a competent lawyer, with rule 3.1-1(k) including the obligation to adapt to changing professional requirements, standards, techniques and practices; rule 6.1-1 puts complete professional responsibility on the lawyer for the practice. British Columbia's Code of Professional Conduct, Chapter 3 covers the same ground for BC lawyers, and the Law Society of Alberta's professional conduct materials for Alberta. If you want the litigation-mechanics background — what a pleading is and what signing one commits you to — Treadstone Law's explanation of pleadings in an Ontario civil lawsuit and of affidavits in Ontario litigation are the plain-language versions.

Questions we get asked

Do I have to declare use of AI in every court?
No. There is no national requirement. Check the practice directions of the court you are filing in; the Federal Court's requirement is the most developed.

Does spellcheck or transcription need declaring in the Federal Court?
No. The Notice expressly excludes AI that only follows pre-set instructions, naming system automation, voice recognition and document editing.

Will declaring hurt my client?
The Court says a Declaration will not in and of itself attract an adverse inference.

See where AI pays off first in your firm.

A 30-minute call is enough to tell you whether AI pays for itself here.