On July 8, 2026, the Immigration and Refugee Board of Canada published a practice notice on the use of artificial intelligence in its proceedings. Chairperson Manon Brassard signed the notice on July 10, 2026. It comes into force September 7, 2026 and applies across all four IRB divisions, the Refugee Protection Division, the Refugee Appeal Division, the Immigration Appeal Division, and the Immigration Division. The Board's stated purpose is to protect the integrity of its processes against the risk that AI tools produce false, fabricated, or inaccurate information.
The notice reaches further than counsel. It binds self-represented claimants, sponsors, and appellants too. This article sets out what the notice requires.
Key Takeaways
- The notice takes effect September 7, 2026 and applies to counsel and self-represented parties alike, across all four IRB divisions.
- Parties remain responsible for anything AI helps produce, including verifying case citations against reliable sources such as CanLII rather than an AI tool's own summaries.
- AI must not be used to draft or materially change a person's own evidence, including basis of claim narratives, affidavits, and witness statements.
- A written declaration, naming the person who verified the content, is required whenever AI generated or materially changed a document filed with the IRB.
- AI use in transcription or translation must always be disclosed, on top of the existing requirement for a human translator's declaration.
- A document that does not comply may be refused, may draw a negative credibility inference, or may lead to a report to counsel's regulator.
01.Why the Board acted
The notice states that AI tools may create false, fabricated, or inaccurate information, and that this risks the integrity of the Board's processes. On the question of submission length, the notice cites a 2025 Federal Court decision, Messa c. Canada (Citoyenneté et Immigration), 2025 CF 1557, at paragraph 11, for the proposition that longer submissions are not better submissions. No official English translation of that decision was available at the time the notice was published. The Board's response is not to prohibit AI outright. It sets expectations of responsibility and disclosure that apply to every party, and a stricter, separate rule for one category of evidence, addressed below.
02.Who the notice covers, and what counts as AI
The notice defines "parties" to include the parties themselves and their counsel, and states this includes self-represented parties. Its definition of AI, set out in the notice's own footnotes, covers systems or technologies that can simulate human-like capabilities such as learning, reasoning, problem-solving, and administrative or managerial decision-making to assist or automate processes, including generative AI, which the notice describes as a subset of AI that generates text, images, code, or other content based on user prompts and training data. That definition is not limited to any one product.
03.Four expectations that apply to every party
The notice sets out four baseline expectations that apply before any document reaches the hard limit on personal evidence, addressed in the next section.
- Take responsibility. A party using AI to build a case must ensure the output is accurate, verifiable, and trustworthy, including systematic verification of legal principles and case citations against reliable sources. The notice states that reliable sources include legal databases, official government websites, commonly referenced commercial publishers, and trusted public services such as CanLII, and that this does not include AI-generated summaries or other AI tools found within those sources.
- Be prepared to respond. A party should expect to answer questions about how AI was used and to establish the authenticity and accuracy of the resulting documents on request.
- Do not pad. Parties should not fill their submissions with AI-generated information that does not relate to the particular facts of their case. Longer submissions are not better submissions.
- Weigh the privacy risk. The notice states that some AI tools do not have sufficient security features to protect information, and directs parties to exercise caution to keep sensitive information confidential.
04.The hard line: personal evidence
The clearest rule in the notice is also the strictest. Parties must not use AI to generate or materially change content that reports or reflects a person's own evidence. The notice names the prohibited uses directly, using AI to draft a basis of claim narrative, affidavit, or witness statement, to rewrite a person's story, or to create or alter photographs, videos, screenshots, or other corroborative evidence.
The notice allows minor assistive functions, spellcheck, grammar correction, and formatting assistance, provided those tools do not generate the evidence or make material changes to it.
05.The declaration
Where a party uses AI to generate or materially change content for a document submitted to the IRB, the notice requires that document to include a declaration, set out in the notice in this form.
“Artificial intelligence (AI) was used to create or materially change text in this document. All content generated by AI, and the authenticity of any cited case law or other legal authorities, has been reviewed and verified by [name].”
Parties may also include in the declaration details about what software was used, what purpose it was used for, and whether AI was used for the entire document or only for particular paragraphs. A declaration is not required for the use of minor assistive functions, spellcheck, grammar correction, or formatting assistance, provided those tools do not generate or materially change the wording of the document. The notice does not define "materially change."
06.Translation and transcription, always disclosed
Use of AI for transcription or translation must always be disclosed. The notice reminds parties that the Rules in all divisions require translated documents to be accompanied by a declaration from a human translator, and that the translator must declare whenever AI was used to assist in translation.
07.What happens if the notice is not followed
The notice sets out the consequences of not following it. If a party does not follow the practice notice, the IRB may refuse to accept the document for filing or otherwise decline to rely on it, draw negative inferences about the credibility of the evidence, disclose information to counsel's professional regulatory authority as outlined in the Board's Policy on Disclosing Information Regarding the Conduct of Authorized Representatives to Regulatory Bodies, restrict or prohibit counsel from appearing before the IRB, or take any other action the Board considers necessary.
For members of the Law Society of Ontario appearing before the IRB, the disclosure route to a regulator, named directly in the notice, means a finding under this notice is not necessarily contained to the file in which it arose.
08.What this means in practice
For a self-represented claimant, the notice is clear on one point above all others. AI should not be used to write, rewrite, or polish a basis of claim narrative, an affidavit, or a witness statement. Minor assistive functions such as spellcheck and formatting remain available.
For counsel, the notice takes effect September 7, 2026, which leaves time to put a verification workflow in place, including a habit of checking citations against CanLII or another primary source rather than a database's AI summary, and a template declaration ready to attach where needed.
Anyone preparing a matter for the IRB after September 7, 2026 should treat disclosure and verification as a filing requirement, not an afterthought. The notice itself, linked at the top of this article, remains the controlling text.