IRB AI Usage Rules: Canada’s Immigration and Refugee Board has put a formal set of rules into force today governing how artificial intelligence can be used by anyone presenting a case before the tribunal, marking one of the clearest lines drawn yet by a Canadian administrative body on AI-generated legal content. The practice notice, titled Use of Artificial Intelligence in IRB Proceedings, took effect September 7, 2026, and applies across all four divisions of the Board, meaning refugee claims, refugee appeals, immigration appeals, and detention or admissibility hearings are all covered.
The move follows growing concern inside the tribunal about AI-generated content appearing in case files. The IRB has previously said it has observed a rise in longer appeal submissions that cite case law which does not actually exist or misrepresents legal precedents, without those submissions becoming any stronger as a result. The new practice notice responds directly to that pattern, setting out what parties and their counsel must do, and must not do, when using AI tools to prepare material for a hearing. We’ll be updating this article monthly as the IRB releases further guidance or enforcement examples under the new rules.
What the New IRB Practice Notice Actually Covers?
The practice notice applies to all parties who communicate with the IRB, a term that includes represented parties, their counsel, and people who represent themselves without a lawyer. It covers content created or materially changed by AI tools, including generative AI systems capable of producing text, images, or other content based on user prompts, and it applies uniformly across the Immigration Division, Immigration Appeal Division, Refugee Protection Division, and Refugee Appeal Division.
The stated purpose of the notice is to protect the integrity of IRB proceedings as AI tools become more widely available to parties preparing submissions, evidence, and legal arguments.
IRB AI Usage Rules Key Facts and Dates
| Detail | Information |
|---|---|
| Practice notice title | Use of Artificial Intelligence in IRB Proceedings |
| Signed by | Manon Brassard, Chairperson, Immigration and Refugee Board of Canada |
| Date signed | July 10, 2026 |
| Effective date | September 7, 2026 |
| Divisions covered | Immigration Division, Immigration Appeal Division, Refugee Protection Division, Refugee Appeal Division |
| Who it applies to | Parties, their counsel, and self-represented individuals |
| Core restriction | AI cannot be used to draft or materially change personal evidence such as narratives, affidavits, or witness statements |
| Disclosure requirement | A specific written declaration is required when AI materially generates or changes submitted content |
| Consequences for non-compliance | Document rejection, negative credibility findings, referral to counsel’s regulatory body, or restrictions on appearing before the IRB |
The Core Rule: Parties Remain Fully Responsible for AI-Generated Content
The central principle running through the entire practice notice is that responsibility for what is submitted to the IRB never shifts to the AI tool itself. Parties using AI to help build a case must ensure that anything the AI produces is accurate, verifiable, and trustworthy, and this includes personally checking that any legal principles or case citations generated by AI are real and correctly represent the law. The IRB has specifically noted that this verification must rely on genuine legal sources such as recognized legal databases, official government websites, established commercial legal publishers, or trusted public resources like CanLII, and explicitly does not accept AI-generated summaries of those sources as a substitute for direct verification.
Parties must also be ready to answer questions about how they used AI in a given proceeding if asked, and must be able to demonstrate the authenticity and accuracy of any AI-assisted documents on request.
IRS Identity Protection PIN: How to Get One and Stop Tax Fraud
Express Entry Labour Shortage Categories 2027: What Canada May Tighten?
AI Cannot Be Used to Create Someone’s Personal Evidence
One of the firmest restrictions in the new rules addresses personal evidence directly. The IRB has made clear that AI must not be used to generate or materially rewrite content that is meant to reflect a person’s own account or experience, including basis of claim narratives, affidavits, and witness statements. These documents must come from the individual’s own knowledge and experience rather than being drafted or substantially reshaped by an AI system.
The notice gives specific examples of prohibited uses, including using AI to draft a basis of claim narrative from scratch, using AI to rewrite a person’s personal story, and using AI to create or alter photographs, videos, screenshots, or other supporting evidence. Ordinary assistive tools such as spellcheck, grammar correction, and formatting help remain acceptable, since these do not generate new content or materially change the substance of what a person has written.
No Padding Submissions With Irrelevant AI-Generated Content
The practice notice also directly addresses a pattern the IRB says it has observed of appeal submissions growing significantly longer without becoming more persuasive. Parties are instructed not to fill their submissions with AI-generated material that does not relate to the specific facts of their case, with the notice stating plainly that longer submissions are not automatically better submissions. This guidance references a 2025 Federal Court decision that touched on the issue of submission length and relevance in the immigration context.
Privacy Considerations When Using AI Tools
Parties are also reminded to think carefully about privacy before feeding sensitive case information into AI tools. The notice cautions that not every AI tool has adequate security features to protect confidential information, meaning personal details, case specifics, or sensitive evidence entered into a general-purpose AI tool could potentially be exposed or retained in ways a party does not intend or control.
When and How Parties Must Declare AI Use
If a party uses AI to generate or materially change content in a document filed with the IRB, the practice notice requires a specific written declaration to be included directly in that document. The declaration must state that AI was used to create or materially change the text, that all AI-generated content and any cited legal authorities have been reviewed and verified, and it must identify the name of the person who carried out that verification. Parties are also permitted, though not required, to add further detail in the declaration, such as which AI software was used, what it was used for, and whether AI was involved in the entire document or only specific sections.
Separately, any use of AI for transcription or translation must always be disclosed, regardless of how minor the AI’s role was. This sits alongside the IRB’s existing rule that translated documents must be accompanied by a declaration from a human translator, and that translator must now specifically state whenever AI was used to assist with the translation.
Importantly, a formal declaration is not required for minor assistive functions like spellcheck, grammar correction, or basic formatting tools, as long as those tools do not generate new content or materially alter the wording of the document.
SSDI Pay Chart: The Least Amount You Can Get Per Month and the Most Common Disabilities
What Happens If Parties Don’t Follow the Rules
The IRB has outlined several possible consequences for parties who fail to comply with the practice notice. The tribunal may refuse to accept a non-compliant document for filing or decline to rely on it in a decision, may draw negative inferences about the credibility of the evidence presented, and may disclose information about the conduct to a lawyer’s or consultant’s professional regulatory body under the IRB’s existing policy on reporting concerns about authorized representatives. In more serious cases, the IRB may restrict or prohibit a specific counsel from appearing before the Board altogether, or take other action it considers necessary to protect the integrity of its proceedings.
How the IRB Itself Uses AI Internally
The new practice notice is focused on parties appearing before the IRB, but the tribunal has also described separate internal rules governing how its own staff and decision-makers may use AI. According to the IRB, any AI use by people working at the Board is governed by internal policies that specifically prohibit using AI to prepare or decide cases, alongside safeguards for how AI may be used in non-adjudicative, administrative work. The IRB has indicated it is exploring AI tools to support faster file preparation and to help decision-makers draft more concise, focused reasons, while maintaining that these tools are intended to streamline the writing process rather than replace human decision-making or influence the substance of a decision.
Why This Practice Notice Was Introduced Now
The timing of this notice follows public reporting earlier in 2026 describing a rise in AI-generated inaccuracies appearing in Canadian immigration and asylum filings. Federal officials had noted that appeal memoranda were becoming noticeably longer without becoming stronger, occasionally citing case law that does not exist or misrepresenting legal precedents in ways that do not support the arguments being made. The new practice notice can be understood as the IRB’s direct, formalized response to that trend, shifting from general awareness of the problem to specific, enforceable expectations for every party appearing before the tribunal.
What Applicants and Counsel Should Do to Comply
- Before filing any AI-assisted document, personally verify every legal citation, case reference, and factual claim the AI produced against a genuine legal source, not an AI-generated summary of that source.
- Never use AI to draft or substantially rewrite a claimant’s personal narrative, affidavit, or witness statement. These must reflect the person’s own account in their own words.
- Include the required written declaration in any document where AI materially generated or changed the content, naming the person who verified it.
- Disclose any use of AI in transcription or translation without exception, and ensure translated documents still carry a human translator’s declaration.
- Avoid using AI simply to lengthen a submission. Focus content specifically on the facts and legal arguments relevant to the individual case.
- Use caution with sensitive case details in general-purpose AI tools, given that not all AI platforms offer adequate data protection.
- Be prepared to explain, if asked by the Board, exactly how and where AI was used in preparing any part of a submission.
UPS Tariff Refund: Who Is Eligible to Get One and How Can You Apply?
Allowed vs Prohibited AI Use at a Glance
| AI use case | Allowed or prohibited | Declaration required |
|---|---|---|
| Spellcheck, grammar correction, formatting | Allowed | No |
| Drafting legal argument or submission text | Allowed with verification | Yes |
| Summarizing case law for internal reference | Allowed with independent verification against original sources | Yes, if included in filed content |
| Drafting a basis of claim narrative | Prohibited | Not applicable, use is barred |
| Rewriting a witness statement or affidavit | Prohibited | Not applicable, use is barred |
| Creating or altering photos, videos, or screenshots as evidence | Prohibited | Not applicable, use is barred |
| Transcription or translation assistance | Allowed | Yes, always |
| Padding a submission with generic AI-generated content | Discouraged and may be treated as non-compliant | Not applicable |
How This Fits a Broader Global Pattern in Legal AI Oversight
The IRB is not acting in isolation. Courts and tribunals in a number of countries have introduced similar guidance in recent years after facing their own instances of AI-generated legal citations that turned out to be fabricated, sometimes called AI hallucinations. Canadian courts, including the Federal Court, have already addressed cases where AI-assisted submissions contained citations to case law that did not exist or misstated what a real precedent actually held. The Messa v. Canada decision referenced in the IRB’s own practice notice is one example of this pattern being addressed directly by a Canadian court in the immigration context.
This broader trend reflects a consistent theme across legal systems adapting to generative AI: the tools themselves are not banned outright, but the responsibility for accuracy is placed squarely on the human party using them, with specific, mandatory verification steps built into the process rather than left to informal best practice.
What This Means for Immigration Consultants and Lawyers Specifically
For licensed immigration consultants and lawyers representing clients before the IRB, the new practice notice adds a concrete compliance obligation on top of existing professional conduct rules. Since the IRB can refer non-compliant conduct to a counsel’s professional regulatory authority, representatives who use AI tools in preparing submissions now have a direct incentive to build verification and declaration steps into their standard workflow rather than treating AI-assisted drafting as a purely internal, undisclosed efficiency tool. This is particularly relevant given the IRB’s own observation that some recent submissions have grown longer while citing legal authorities that do not hold up under scrutiny, a pattern the Board has directly linked to unverified AI use.
Self-represented individuals, who are also covered by the notice, face the same expectations without the benefit of professional training in legal research or citation verification, which makes the availability of reliable, human-reviewed information about the rules especially important for that group.
Timeline for Compliance
There is no phased grace period built into the practice notice itself. It became effective on September 7, 2026, meaning the expectations described above apply immediately to documents and submissions filed with the IRB from that date forward, across all four divisions of the Board.
Official Resources
| Resource | Link |
|---|---|
| IRB Practice Notice on AI Use | https://www.irb-cisr.gc.ca/en/legal-policy/procedures/Pages/practice-notice-ai-proceedings.aspx |
| IRB Procedures and Practice Notices | https://www.irb-cisr.gc.ca/en/legal-policy/procedures/pages/index.aspx |
| My Case (IRB secure portal) | https://mycase-mondossier.irb-cisr.canada.ca/en-US/ |
| IRB Policy on Disclosing Information to Regulatory Bodies | https://www.irb-cisr.gc.ca/en/legal-policy/policies/Pages/PolCondRep.aspx |
FAQs About IRB AI Usage Rules
When did the IRB’s new AI usage rules take effect?
The practice notice came into force on September 7, 2026, and applies immediately to all documents and submissions filed with the Board from that date onward.
Can I use AI to write my refugee claim narrative?
No. The practice notice specifically prohibits using AI to draft or materially rewrite a basis of claim narrative, affidavit, or witness statement. These must be based entirely on the person’s own knowledge and experience.
Do I have to disclose AI use every time I use a writing tool?
Not for minor assistive functions like spellcheck, grammar correction, or formatting. A formal written declaration is only required when AI generates new content or materially changes the wording of a document.
What happens if I don’t follow the IRB’s AI usage rules?
The IRB can refuse to accept the document, draw negative inferences about the credibility of the evidence, report the conduct to a regulatory body, or restrict a lawyer or consultant from appearing before the Board.
Does this rule apply to translations done with AI assistance?
Yes. Any use of AI in transcription or translation must always be disclosed, and translated documents must still include a declaration from a human translator confirming whether AI was used.
Does the IRB use AI to make decisions on cases?
According to the IRB, internal policy currently prohibits using AI to prepare or decide cases. AI use inside the Board is limited to non-adjudicative, administrative functions such as drafting assistance, with human decision-makers retaining full control over the substance of decisions.
Who does the new practice notice apply to?
It applies to all parties appearing before the IRB, including represented parties, their counsel, and people who represent themselves without a lawyer, across all four divisions of the Board.
Conclusion
The IRB’s new practice notice on AI usage represents a direct, formalized response to rising concerns about fabricated case law and unreliable AI-generated content appearing in immigration and refugee proceedings. Effective September 7, 2026, it draws a clear line between acceptable assistive use of AI and the kind of AI-generated content that risks undermining the integrity of a case, particularly personal narratives and evidence that must come directly from the individual involved. Parties and counsel who rely on AI tools going forward should build verification and disclosure into their standard process now, since the consequences for non-compliance range from a rejected document to a referral to a professional regulatory body.
USCIS Addis Ababa Field Office: New International Office Opens September 9
Is It Hard to Refinance Student Loan, and What Are the Pros and Cons of Doing It?
