AI can make a legal argument easier to generate. It can also make the argument harder - and more expensive - for a court to process.
That is the institutional concern emerging from 1182573 B.C. Ltd. v. Gerber, 2026 BCCA 399. Justice Susan Griffin addressed AI-assisted submissions after a self-represented party disclosed using the technology. Her observations went beyond fabricated authorities: irrelevant citations, excessive length, incomprehensible legal language and aggressive tone can consume judicial attention without advancing a case. Canadian Lawyer
For Canadian lawyers, the important shift is from individual accuracy to collective capacity. A document may burden the justice system even when its citations are real.
The problem extends beyond hallucinations
Much of the discussion about legal AI has focused on nonexistent cases. That remains a serious concern, but citation verification answers only one question: does the authority exist?
Other questions require legal judgment. Does the case support the proposition? Does the proposition matter to the application? Is the requested relief available? Has the submission addressed the governing test?
An argument can pass the first check and fail all the others.
Griffin’s observations, as reported by Canadian Lawyer, did not identify a particular AI-generated error in the submissions. That distinction matters: the decision should not be presented as another finding that a litigant filed fabricated authorities. Canadian Lawyer
Judicial attention is part of access to justice
Generating additional pages has become cheap. Reviewing them remains human work.
A judge must identify the material issues, understand the evidence and decide which arguments require an answer. Unfocused submissions can complicate that task for everyone involved.
The resulting access-to-justice issue is broader than the litigant who used the tool. Time spent untangling an irrelevant argument is time unavailable for another matter.
This is an editorial implication of the warning, rather than a finding that AI is responsible for court delays generally.
Self-represented litigants need usable guidance
The response cannot rest entirely on telling people to stop using technology.
A person unable to obtain counsel may use AI because they need help understanding a process, organizing documents or explaining their position. Those are understandable needs.
The challenge is to provide guidance that distinguishes useful assistance from persuasive-looking output that leads a litigant away from the actual issue.
Court forms, focused procedural explanations and examples of relevant submissions may help people understand what a court needs from them.
What lawyers should take from the warning
AI-assisted filing review should include relevance, proportionality and tone alongside citation checks.
Counsel can ask four questions before filing: What decision is the court being asked to make? Which legal test governs it? What evidence addresses that test? Which passages can be removed without weakening the submission?
The same discipline applies whether the first draft came from a lawyer, a client or software.
The value of a submission lies in the judicial decision it helps the court make.
Related coverage: read our analysis of AI errors and junior-lawyer training.
Sources: Canadian Lawyer.