Law360 Canada (July 21, 2026, 11:10 AM EDT) --

Tom Macintosh Zheng

By now, the problem is well-known. In March, we reported in Law360 Canada that Canadian courts and tribunals had flagged fictitious case citations in more than 111 decisions, a number that has reached 186 as of early July. In a June article, we showed that the consequences for the people who file them are inconsistent. This article asks the third and final question: what are the courts and tribunals themselves doing about it?

To answer that, we searched for AI policies across 249 Canadian courts and tribunals.

The headline numbers

As of the date of this article:

• 167 (67.1 per cent) have no policies.
• 82 (32.9 per cent) have adopted dedicated AI rules or guidance for parties appearing before them.
• 19 (7.6 per cent) require a party to disclose their use of AI.
• 6 (2.4 per cent) prohibit some use of AI.

Put differently: at more than 90 per cent of Canada’s courts and tribunals we are tracking, a party can use AI to prepare their entire case and never has to tell anyone.

Four categories of policies

Warning: The largest category teaches litigants to be cautious. Superior courts in Alberta, Quebec, Nova Scotia and Newfoundland and Labrador, and tribunals such as the Canadian Human Rights Tribunal and Tribunals Ontario, all deliver the same three messages: be cautious with AI-derived material, verify citations against authoritative sources, and remember that you are responsible for what you file.

Disclose: Nineteen bodies require parties to reveal their use of AI. This approach began in June 2023, when the Court of King’s Bench of Manitoba and the Supreme Court of Yukon each directed parties to state how AI was used in materials before the court. The Federal Court followed suit in December 2023, requiring a declaration in the first paragraph of any document containing AI-generated content. The Trademarks Opposition Board, the Canada Industrial Relations Board and the Supreme Court of the Northwest Territories have since adopted versions of that declaration, and Quebec’s Administrative Labour Tribunal and Financial Markets Administrative Tribunal followed with their own disclosure requirements in late 2025.

The depth of disclosure varies by body. The Saskatchewan Labour Relations Board asks for a single sentence stating that AI was used. The Alberta Utilities Commission requires more: a declaration identifying the paragraphs in which AI was used and explaining how it was used. At B.C.’s Provincial Court, disclosure is owed when the judge asks for it.

Verify: A third approach targets the same problem from a different angle: requiring parties to attest that they verified their authorities. Ontario’s Rules of Civil Procedure now require the familiar authenticity certification in every factum, whether or not AI was involved. Others combine verification with disclosure. The Supreme Court of Yukon, which has required disclosure since 2023, added a certification obligation in 2026. The Ontario Land Tribunal builds both into a single declaration: the filer discloses that AI was used and confirms that every authority cited has been reviewed and verified.

Prohibit: Six bodies ban some use of AI outright. While none of them prohibits using AI for legal research, what they prohibit is what goes into AI and what comes out of it. The Canada Industrial Relations Board and the Manitoba Labour Board bar AI-generated witness statements and affidavits. B.C.’s Civil Resolution Tribunal bars parties from filing AI-created or altered evidence. The B.C. Employment Standards Tribunal prohibits fabricated records and simulated testimony. On the input side, B.C.’s Health Professions Review Board forbids putting any part of the record into an AI tool, and the Canadian International Trade Tribunal bars counsel from entering third-party confidential information.

What a good policy looks like

For the two-thirds of Canadian courts and tribunals without a policy, the good news is they do not need to invent anything. The best elements already exist, and they only need to be combined.

First, require disclosure, but indicate that such disclosure will not attract an adverse inference. A party who admits to using AI must know that the court will not hold using AI against them. Otherwise, some litigants may choose to stay silent.

Second, add guidance that teaches instead of simply warns. The Social Security Tribunal tells parties, in plain language, to search for the case on legal databases, and that if they cannot find it, it may not exist. It is not enough to tell people to verify the record. Many self-represented litigants do not know where the correct information lives. Telling someone to verify is not the same as showing them how.

Finally, give people a way out. B.C.’s Property Assessment Appeal Board tells parties to correct the record as soon as they discover an error. Every policy describes what happens when you get caught. Almost none describes what to do when you catch an error yourself.

All three share the same underlying idea: a party who tells the truth about AI should never be worse off than one who stays quiet.

Tom Macintosh Zheng is a Toronto-based former commercial litigator. He is now building online tools to help Canadians access and understand our justice system. To see the full database tracking AI policies in Canadian courts and tribunals, which is updated daily, click here.

The opinions expressed are those of the author(s) and do not necessarily reflect the views of the author’s firm, its clients, LexisNexis Canada, Law360 Canada, or any of its or their respective affiliates. This article is for general information purposes and is not intended to be and should not be taken as legal advice.

Interested in writing for us? To learn more about how you can add your voice to Law360 Canada, contact Analysis Editor Richard Skinulis at Richard.Skinulis@lexisnexis.ca or call 437-828-6772.